diff --git a/frontend/src/App.tsx b/frontend/src/App.tsx
index 6597018..547cedf 100644
--- a/frontend/src/App.tsx
+++ b/frontend/src/App.tsx
@@ -49,9 +49,11 @@ import {
 import {
   getAudioPlaybackState,
   pauseAudioAsset,
+  seekAudioAsset,
   subscribeAudioPlayback,
   toggleAudioAsset,
   useAudioPlayback,
+  useAudioProgress,
 } from "./audioPlayback";
 import {
   hasLoadedCardPreview,
@@ -535,7 +537,14 @@ function ruleFieldName(ruleField: string | null) {
   return fields[fields.length - 1] || "动画引用";
 }
 
+const mediaCategoryKinds: Record<string, MediaKind> = {
+  voices: "voice",
+  sounds: "sound",
+  music: "music",
+};
+
 const mediaGroupLabels: Record<string, string> = {
+  theme_music: "主题音乐",
   selection_voice: "选中回应",
   movement_voice: "移动指令",
   combat_voice: "攻击指令",
@@ -730,7 +739,7 @@ function staticBuildInfo(currentVersion: string) {
 }
 const previewAngleOptions: Array<{ value: PreviewAngle; label: string }> = [
   { value: 0, label: "正面" },
-  { value: 1, label: "右前侧（推荐）" },
+  { value: 1, label: "右前侧" },
   { value: 2, label: "右侧" },
   { value: 3, label: "右后侧" },
   { value: 4, label: "背面" },
@@ -782,7 +791,7 @@ type BrowsingLocation = {
 
 const audioFormats = ["bag_audio", "wav", "aud"];
 const imageFormats = ["shp", "tmp", "pcx", "pal", "map"];
-const defaultVisibleFormats = ["bag_audio", "wav", "aud"];
+const defaultVisibleFormats = ["bag_audio", "wav", "aud", "shp", "vxl", "hva", "pal", "vpl", "video"];
 
 function preloadEntityAudioResources(entity: GameEntity) {
   const primaryUrls: string[] = [];
@@ -817,8 +826,13 @@ const assetCategories: Array<{
 }> = [
   { id: "voices", label: "游戏语音", formats: ["bag_audio", "wav", "aud"], icon: "play" },
   { id: "sounds", label: "游戏音效", formats: ["bag_audio", "wav", "aud"], icon: "play" },
+  { id: "music", label: "游戏音乐", formats: ["wav", "aud"], icon: "sound" },
   { id: "maps", label: "地图", formats: ["map"], icon: "grid" },
   { id: "images", label: "图像", formats: ["pcx"], icon: "image" },
+  { id: "sprites", label: "精灵动画", formats: ["shp"], icon: "image" },
+  { id: "models", label: "三维模型", formats: ["vxl", "hva"], icon: "unit" },
+  { id: "palettes", label: "调色板", formats: ["pal", "vpl"], icon: "swatch" },
+  { id: "videos", label: "过场影片", formats: ["video"], icon: "play" },
   { id: "terrain", label: "地形素材", formats: ["tmp"], icon: "image" },
   { id: "rules", label: "规则文本", formats: ["ini", "csf", "text"], icon: "file" },
 ];
@@ -1049,6 +1063,17 @@ function ruleColumnSpan(label: string, value: string) {
   return width > 72 ? 3 : width > 34 ? 2 : 1;
 }
 
+// Clamp spans against the 3-column cursor so every visual row is filled;
+// otherwise a wide cell jumps to the next grid row and leaves a blank hole.
+function ruleSpanLayout(rows: Array<[string, string]>, labels: Record<string, string>) {
+  let cursor = 0;
+  return rows.map(([key, value]) => {
+    const span = Math.min(ruleColumnSpan(labels[key], value), 3 - cursor);
+    cursor = (cursor + span) % 3;
+    return span;
+  });
+}
+
 const appliedRememberedScrollResets = new Map<string, number>();
 
 function useRememberedScroll<
@@ -1444,8 +1469,8 @@ function ExplorerApp() {
     || visibleCategories[0]
     || null;
   const selectedCategoryId = selectedCategory?.id || "";
-  const isMediaCategory = selectedCategoryId === "voices" || selectedCategoryId === "sounds";
-  const mediaKind: MediaKind = selectedCategoryId === "voices" ? "voice" : "sound";
+  const isMediaCategory = selectedCategoryId in mediaCategoryKinds;
+  const mediaKind: MediaKind = mediaCategoryKinds[selectedCategoryId] || "sound";
   const categoryFormats = (selectedCategory?.formats || [])
     .filter((formatName) => enabledFormats.includes(formatName));
   const assetFormats = assetFormatTag && categoryFormats.includes(assetFormatTag)
@@ -2697,16 +2722,16 @@ function ExplorerApp() {
                   })}
                 </div>
               </section>
-              {visibleCategories.some((item) => ["voices", "sounds"].includes(item.id)) && <section className="tree-branch media-tree-branch">
+              {visibleCategories.some((item) => item.id in mediaCategoryKinds) && <section className="tree-branch media-tree-branch">
                 <div className="tree-parent"><Icon name="play" /><strong>声音</strong><em>{mediaKindCounts.reduce((count, item) => count + item.count, 0) || categoryCount(stats, audioFormats)}</em></div>
                 <div className="tree-children">
-                  {visibleCategories.filter((item) => ["voices", "sounds"].includes(item.id)).map((item) => <div className="tree-child-group" key={item.id}>
-                    <button title={item.label} className={view === "assets" && assetCategory === item.id && !mediaGroup ? "active" : ""} onClick={() => selectAssetCategory(item.id)}><span><Icon name={item.id === "voices" ? "voice" : "sound"} /><b>{item.label}</b></span>{mediaKindCounts.length > 0 && <em>{mediaKindCounts.find((count) => count.kind === (item.id === "voices" ? "voice" : "sound"))?.count || 0}</em>}</button>
-                    {view === "assets" && assetCategory === item.id && mediaGroups.filter((group) => group.group.endsWith(item.id === "voices" ? "_voice" : "_sound")).map((group) => <button title={mediaGroupLabels[group.group] || group.group} className={`tree-grandchild ${mediaGroup === group.group ? "active" : ""}`} key={group.group} onClick={() => selectMediaGroup(mediaGroup === group.group ? "" : group.group)}><span><Icon name={item.id === "voices" ? "voice" : "sound"} /><b>{mediaGroupLabels[group.group] || group.group}</b></span><em>{group.count}</em></button>)}
-                  </div>)}
+                  {visibleCategories.filter((item) => item.id in mediaCategoryKinds).map((item) => { const kind = mediaCategoryKinds[item.id]; return <div className="tree-child-group" key={item.id}>
+                    <button title={item.label} className={view === "assets" && assetCategory === item.id && !mediaGroup ? "active" : ""} onClick={() => selectAssetCategory(item.id)}><span><Icon name={item.id === "voices" ? "voice" : "sound"} /><b>{item.label}</b></span>{mediaKindCounts.length > 0 && <em>{mediaKindCounts.find((count) => count.kind === kind)?.count || 0}</em>}</button>
+                    {view === "assets" && assetCategory === item.id && mediaGroups.filter((group) => group.group.endsWith(`_${kind}`)).map((group) => <button title={mediaGroupLabels[group.group] || group.group} className={`tree-grandchild ${mediaGroup === group.group ? "active" : ""}`} key={group.group} onClick={() => selectMediaGroup(mediaGroup === group.group ? "" : group.group)}><span><Icon name={item.id === "voices" ? "voice" : "sound"} /><b>{mediaGroupLabels[group.group] || group.group}</b></span><em>{group.count}</em></button>)}
+                  </div>; })}
                 </div>
               </section>}
-              {visibleCategories.filter((item) => !["voices", "sounds"].includes(item.id)).map((item) => (
+              {visibleCategories.filter((item) => !(item.id in mediaCategoryKinds)).map((item) => (
                 <button key={item.id} title={item.label} className={`tree-leaf ${view === "assets" && assetCategory === item.id ? "active" : ""}`} onClick={() => selectAssetCategory(item.id)}>
                   <span><Icon name={item.icon} /><b>{item.label}</b></span><em>{categoryCount(stats, item.formats.filter((formatName) => enabledFormats.includes(formatName)))}</em>
                 </button>
@@ -2936,9 +2961,10 @@ function AssetGridCard({ asset, selected, onSelect }: { asset: Asset; selected:
   return (
     <button className={`asset-card ${selected ? "selected" : ""}`} onClick={() => onSelect(asset.id)}>
       <span className={`asset-card-preview format-${asset.format}`}>
-        {hasThumbnail ? <img loading="lazy" src={api.previewUrl(asset.id, 0, "", 3)} alt="" onError={(event) => { event.currentTarget.hidden = true; }} />
-          : isAudio ? <span className="audio-glyph" aria-hidden="true">{[4, 11, 7, 17, 12, 20, 9, 14, 5, 10].map((height, index) => <i key={index} style={{ height }} />)}</span>
-            : <Icon name={assetIcon(asset.format)} size={32} />}
+        {asset.format === "video" ? <img loading="lazy" src={api.videoPosterUrl(asset.id)} alt="" onError={(event) => { event.currentTarget.hidden = true; }} />
+          : hasThumbnail ? <img loading="lazy" src={api.previewUrl(asset.id, 0, "", 3)} alt="" onError={(event) => { event.currentTarget.hidden = true; }} />
+            : isAudio ? <span className="audio-glyph" aria-hidden="true">{[4, 11, 7, 17, 12, 20, 9, 14, 5, 10].map((height, index) => <i key={index} style={{ height }} />)}</span>
+              : <Icon name={assetIcon(asset.format)} size={32} />}
       </span>
       <span className="asset-card-copy"><strong title={assetDisplayName(asset)}>{assetDisplayName(asset)}</strong><small>{formatLabels[asset.format] || asset.format.toUpperCase()} · {formatBytes(asset.size)}</small></span>
     </button>
@@ -4499,6 +4525,30 @@ function CompactAudioPlayer({ assetId, label }: {
   </span>;
 }
 
+// Playhead and elapsed/total readout for the shared audio element. The bar
+// stays laid out even before playback so the row never jumps, but it only
+// becomes seekable once this asset owns playback and its duration is known.
+function AudioProgress({ assetId, durationSeconds }: {
+  assetId: string;
+  durationSeconds?: number;
+}) {
+  const progress = useAudioProgress();
+  const owned = progress.assetId === assetId && progress.duration > 0;
+  const currentTime = owned ? progress.currentTime : 0;
+  const duration = owned ? progress.duration : durationSeconds || 0;
+  const played = duration > 0 ? (currentTime / duration) * 100 : 0;
+  return <>
+    <input type="range" className="audio-progress" min={0} max={duration || 1}
+      step={Math.min(Math.max((duration || 1) / 100, 0.01), 1)}
+      value={currentTime} disabled={!owned}
+      onChange={(event) => seekAudioAsset(Number(event.target.value))}
+      style={{ "--audio-progress": `${played}%` } as CSSProperties}
+      title={owned ? "拖动以跳转" : "播放后可跳转"}
+      aria-label="播放进度" aria-valuetext={formatDuration(currentTime)} />
+    <span className="audio-progress-time">{formatDuration(currentTime)} <i>/</i> {duration > 0 ? formatDuration(duration) : "--:--"}</span>
+  </>;
+}
+
 function AudioDownloadAction({ assetId, label }: { assetId: string; label: string }) {
   return <a className="audio-download-action" href={api.contentUrl(assetId)} download={isStaticSnapshot ? `${label}.ogg` : true} title={`下载 ${label}`} aria-label={`下载 ${label}`}><Icon name="download" size={15} /></a>;
 }
@@ -4955,11 +5005,12 @@ function EntitySoundSample({ sample, voiceTextPreference }: { sample: MediaSampl
   );
   const localizedText = preferredText?.value !== originalText ? preferredText : null;
   const internalName = audioDisplayName(sample.name);
-  return <div className="media-sample">
-    <strong className="media-sample-id" title={sample.asset?.display_name || sample.name}>{internalName}{sample.weight > 1 && <em>×{sample.weight}</em>}</strong>
+  const missing = !sample.asset || !audioFormats.includes(sample.asset.format);
+  return <div className={missing ? "media-sample media-sample-no-audio" : "media-sample"}>
+    <strong className="media-sample-id" title={sample.asset?.display_name || sample.name}>{internalName}{sample.weight > 1 && <em>×{sample.weight}</em>}{missing && <em className="media-sample-missing">未解析</em>}</strong>
     {sample.asset && audioFormats.includes(sample.asset.format)
       ? <CompactAudioPlayer assetId={sample.asset.id} label={internalName} />
-      : <em className="media-sample-missing">未解析</em>}
+      : null}
     <span className={`media-sample-texts ${localizedText ? "bilingual" : "single"}`}>
       {originalText && <span className="media-sample-copy" title={originalText}><b>原文</b>{originalText}</span>}
       {localizedText && <span className="media-sample-copy" title={localizedText.value}><b>{localizedText.label}</b>{localizedText.value}</span>}
@@ -5437,6 +5488,7 @@ function EntityDetailPanel({ sourceId, sourceRevision = "", entity, loading, pla
   if (loading && !entity) return <aside className="detail-panel panel empty-detail"><div className="radar small"><span /></div><strong>正在读取单位详情…</strong></aside>;
   if (!entity) return <aside className="detail-panel panel empty-detail"><div className="empty-detail-icon"><Icon name="unit" size={30} /></div><strong>选择单位</strong></aside>;
   const rules = Object.entries(entity.rules).filter(([key]) => ruleLabels[key]);
+  const ruleSpans = ruleSpanLayout(rules, ruleLabels);
   const soundAssociations = mergeSoundAssociations(entity.media);
   const soundCount = soundAssociations.reduce((count, association) => count + association.samples.length, 0);
   const hasBodySequences = bodyAnimationAssociations.length > 0;
@@ -5666,7 +5718,7 @@ function EntityDetailPanel({ sourceId, sourceRevision = "", entity, loading, pla
 
             {rules.length > 0 && <details className="entity-section compact-section entity-rules" open>
               <summary><span>规则属性</span><em>{rules.length}</em></summary>
-              <div className="metadata"><dl>{rules.map(([key, value]) => <div className={`rule-span-${ruleColumnSpan(ruleLabels[key], value)}`} key={key}><dt>{ruleLabels[key]}</dt><dd>{value}</dd></div>)}</dl></div>
+              <div className="metadata"><dl>{rules.map(([key, value], index) => <div className={`rule-span-${ruleSpans[index]}`} key={key}><dt>{ruleLabels[key]}</dt><dd>{value}</dd></div>)}</dl></div>
             </details>}
 
             {dependencyGroups.length > 0 && <div className="entity-dependencies">
@@ -5874,16 +5926,16 @@ function DetailPanel({ asset, metadata, textAsset, textQuery, setTextQuery, fram
         <div className="detail-heading-line"><span className="format-pill">{formatLabels[asset.format] || asset.format.toUpperCase()}</span><h2 title={detailHeading}>{detailHeading}</h2></div>
         <div className="detail-actions">
           {onPopout && <button type="button" className="icon-button" onClick={onPopout} title="在独立窗口中打开" aria-label="在独立窗口中打开"><Icon name="popout" /></button>}
-          {isAudio
-            ? <div className="audio-header-player">
-              <CompactAudioPlayer assetId={asset.id} label={assetDisplayName(asset)} />
-              <span className="audio-header-meta" title={metadata?.audio_codec || ""}>{metadata?.duration_seconds !== undefined ? formatDuration(metadata.duration_seconds) : ""}</span>
-              <AudioDownloadAction assetId={asset.id} label={assetDisplayName(asset)} />
-            </div>
-            : <a className="icon-button" href={api.contentUrl(asset.id)} title="导出原始文件" aria-label="导出原始文件"><Icon name="download" /></a>}
+          {!isAudio && <a className="icon-button" href={api.contentUrl(asset.id)} title="导出原始文件" aria-label="导出原始文件"><Icon name="download" /></a>}
         </div>
       </div>
 
+      {isAudio && <div className="audio-detail-player">
+        <CompactAudioPlayer assetId={asset.id} label={assetDisplayName(asset)} />
+        <AudioProgress assetId={asset.id} durationSeconds={metadata?.duration_seconds} />
+        <AudioDownloadAction assetId={asset.id} label={assetDisplayName(asset)} />
+      </div>}
+
       {hasAudioRelationshipTabs && <div className="entity-detail-tabs asset-detail-tabs" role="tablist" aria-label="声音详细信息">
         <button type="button" role="tab" id="audio-associations-tab" aria-controls="audio-associations-panel" aria-selected={activeAudioDetailTab === "associations"} className={activeAudioDetailTab === "associations" ? "active" : ""} onClick={() => selectAudioDetailTab("associations")}>关联 <em>{audioRelationshipItems.length}</em></button>
         <button type="button" role="tab" id="audio-data-tab" aria-controls="audio-data-panel" aria-selected={activeAudioDetailTab === "data"} className={activeAudioDetailTab === "data" ? "active" : ""} onClick={() => selectAudioDetailTab("data")}>数据 <em>{audioDataCount}</em></button>
@@ -5916,8 +5968,11 @@ function DetailPanel({ asset, metadata, textAsset, textQuery, setTextQuery, fram
 
       {asset.format === "video" && <div className="video-preview">
         {videoRequested && !videoFailed
-          ? <video controls autoPlay preload="metadata" src={api.videoUrl(asset.id)} onLoadedData={() => setVideoFailed(false)} onError={() => setVideoFailed(true)}>浏览器不支持视频播放。</video>
-          : <button type="button" className="button primary" onClick={() => { setVideoFailed(false); setVideoRequested(true); }}><Icon name="play" />{videoFailed ? "重试转换" : "转换并播放"}</button>}
+          ? <video controls autoPlay preload="metadata" src={api.videoUrl(asset.id)} poster={api.videoPosterUrl(asset.id)} onLoadedData={() => setVideoFailed(false)} onError={() => setVideoFailed(true)}>浏览器不支持视频播放。</video>
+          : <>
+            <img className="video-poster" src={api.videoPosterUrl(asset.id)} alt="" onError={(event) => { event.currentTarget.hidden = true; }} />
+            <button type="button" className="video-convert" onClick={() => { setVideoFailed(false); setVideoRequested(true); }} title={videoFailed ? "重试转换" : "转换并播放"} aria-label={videoFailed ? "重试转换" : "转换并播放"}><Icon name="play" size={22} /></button>
+          </>}
         {videoFailed && <strong className="video-error">视频转换失败</strong>}
       </div>}
 
diff --git a/frontend/src/api.ts b/frontend/src/api.ts
index 1cfa878..11ade81 100644
--- a/frontend/src/api.ts
+++ b/frontend/src/api.ts
@@ -383,7 +383,7 @@ export interface CountryFacet {
   count: number;
 }
 
-export type MediaKind = "voice" | "sound" | "unknown";
+export type MediaKind = "voice" | "sound" | "music" | "unknown";
 export type MediaSort = "name_asc" | "name_desc" | "description_asc";
 
 export interface MediaItem {
@@ -678,6 +678,7 @@ export const api = {
     ? staticAudioUrl(assetId)
     : `/api/assets/${assetId}/media`,
   videoUrl: (assetId: string) => `/api/assets/${assetId}/video.mp4`,
+  videoPosterUrl: (assetId: string) => `/api/assets/${assetId}/poster.png`,
   entityPreviewUrl: (
     sourceId: string,
     entityId: string,
diff --git a/frontend/src/audioPlayback.ts b/frontend/src/audioPlayback.ts
index 19ceec8..7d10d54 100644
--- a/frontend/src/audioPlayback.ts
+++ b/frontend/src/audioPlayback.ts
@@ -8,8 +8,16 @@ interface AudioPlaybackState {
   loading: boolean;
 }
 
+interface AudioProgressState {
+  assetId: string;
+  currentTime: number;
+  duration: number;
+}
+
 const listeners = new Set<() => void>();
+const progressListeners = new Set<() => void>();
 let playbackState: AudioPlaybackState = { assetId: "", playing: false, loading: false };
+let progressState: AudioProgressState = { assetId: "", currentTime: 0, duration: 0 };
 let sharedAudio: HTMLAudioElement | null = null;
 let sourceUrl = "";
 
@@ -23,6 +31,27 @@ function publish(next: AudioPlaybackState) {
   for (const listener of listeners) listener();
 }
 
+// Progress lives in its own channel so the frequent timeupdate tick only
+// re-renders the progress bar instead of every play button on screen.
+function publishProgress(next: AudioProgressState) {
+  if (
+    next.assetId === progressState.assetId
+    && next.currentTime === progressState.currentTime
+    && next.duration === progressState.duration
+  ) return;
+  progressState = next;
+  for (const listener of progressListeners) listener();
+}
+
+function readProgress() {
+  if (!sharedAudio) return;
+  publishProgress({
+    assetId: playbackState.assetId,
+    currentTime: sharedAudio.currentTime,
+    duration: Number.isFinite(sharedAudio.duration) ? sharedAudio.duration : 0,
+  });
+}
+
 function audioElement() {
   if (sharedAudio) return sharedAudio;
   const audio = new Audio();
@@ -33,6 +62,9 @@ function audioElement() {
   audio.addEventListener("pause", () => publish({ assetId: playbackState.assetId, playing: false, loading: false }));
   audio.addEventListener("ended", () => publish({ assetId: playbackState.assetId, playing: false, loading: false }));
   audio.addEventListener("error", () => publish({ assetId: playbackState.assetId, playing: false, loading: false }));
+  audio.addEventListener("timeupdate", readProgress);
+  audio.addEventListener("durationchange", readProgress);
+  audio.addEventListener("seeked", readProgress);
   sharedAudio = audio;
   return audio;
 }
@@ -52,6 +84,21 @@ export function useAudioPlayback() {
   return useSyncExternalStore(subscribeAudioPlayback, getAudioPlaybackState, getAudioPlaybackState);
 }
 
+export function getAudioProgress() {
+  return progressState;
+}
+
+export function subscribeAudioProgress(listener: () => void) {
+  progressListeners.add(listener);
+  return () => {
+    progressListeners.delete(listener);
+  };
+}
+
+export function useAudioProgress() {
+  return useSyncExternalStore(subscribeAudioProgress, getAudioProgress, getAudioProgress);
+}
+
 export function playAudioAsset(assetId: string, url: string) {
   const audio = audioElement();
   const cachedUrl = cachedAudioResourceUrl(url);
@@ -61,6 +108,7 @@ export function playAudioAsset(assetId: string, url: string) {
     audio.pause();
     sourceUrl = playableUrl;
     publish({ assetId, playing: false, loading: true });
+    publishProgress({ assetId, currentTime: 0, duration: 0 });
     audio.src = playableUrl;
     audio.preload = "auto";
   }
@@ -74,6 +122,12 @@ export function pauseAudioAsset(assetId?: string) {
   sharedAudio?.pause();
 }
 
+export function seekAudioAsset(seconds: number) {
+  if (!sharedAudio || !Number.isFinite(sharedAudio.duration)) return;
+  sharedAudio.currentTime = Math.min(Math.max(seconds, 0), sharedAudio.duration);
+  readProgress();
+}
+
 export function toggleAudioAsset(assetId: string, url: string) {
   const audio = audioElement();
   if (playbackState.assetId === assetId && (!audio.paused || playbackState.loading)) {
diff --git a/frontend/src/styles.css b/frontend/src/styles.css
index 83772e3..a7ac661 100644
--- a/frontend/src/styles.css
+++ b/frontend/src/styles.css
@@ -1,6 +1,6 @@
 :root {
   color-scheme: dark;
-  font-family: "Segoe UI", "Microsoft YaHei UI", "Microsoft YaHei", sans-serif;
+  font-family: "Segoe UI", -apple-system, "PingFang SC", "Hiragino Sans GB", "Microsoft YaHei UI", "Microsoft YaHei", sans-serif;
   color: #e8e9eb;
   background: #101114;
   font-synthesis: none;
@@ -20,7 +20,8 @@
   --entity-accent: #74a9d4;
   --voice-accent: #c18bd6;
   --sound-accent: #d6ad5d;
-  --sans-condensed: "Bahnschrift SemiCondensed", "Arial Narrow", "Microsoft YaHei UI", sans-serif;
+  --sans-condensed: "Bahnschrift SemiCondensed", "Avenir Next Condensed", "Arial Narrow", "PingFang SC", "Microsoft YaHei UI", sans-serif;
+  --mono: "Cascadia Mono", Consolas, "SF Mono", Menlo, Monaco, monospace;
 }
 
 * { box-sizing: border-box; }
@@ -45,7 +46,7 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .button.ghost { background: transparent; }
 .button.compact { min-height: 34px; padding: 0 12px; font-size: 12px; }
 .button.large { min-height: 48px; padding: 0 21px; font-size: 14px; }
-.button:disabled { opacity: .45; cursor: wait; }
+.button:disabled { opacity: .45; cursor: not-allowed; }
 
 .workspace { --sidebar-width: 224px; --detail-panel-size: 430px; height: 100vh; min-height: 0; display: grid; }
 .workspace.detail-right { grid-template-columns: var(--sidebar-width) minmax(360px, 1fr) 6px minmax(320px, var(--detail-panel-size)); }
@@ -84,9 +85,9 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .sidebar-settings:hover { border-color: #555a64; color: #f0f1f2; background: #1d2025; }
 .sidebar-settings .icon { color: #d05a54; }
 .sidebar-settings > i { width: 7px; height: 7px; margin-left: auto; border-radius: 50%; background: #e1534d; box-shadow: 0 0 8px rgba(225,83,77,.7); }
-.source-heading { padding: 24px 17px 17px; border-bottom: 1px solid var(--line-soft); }
+.source-heading { padding: 17px; border-bottom: 1px solid var(--line-soft); }
 .eyebrow { display: block; color: #a7abb3; font: 600 10px/1 var(--sans-condensed); letter-spacing: .18em; text-transform: uppercase; }
-.source-select-wrap { position: relative; display: block; margin-top: 10px; }
+.source-select-wrap { position: relative; display: block; }
 .source-select-wrap select { width: 100%; padding: 0 24px 0 26px; border: 0; color: #f0f0f1; background: transparent; font-size: 15px; font-weight: 600; appearance: none; cursor: pointer; }
 .source-select-wrap .icon { position: absolute; right: 0; top: 3px; color: var(--quiet); transform: rotate(90deg); pointer-events: none; }
 .source-select-wrap > .icon:first-child { right: auto; left: 0; transform: none; }
@@ -105,7 +106,7 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .library-tree button { position: relative; width: 100%; height: 38px; display: flex; align-items: center; justify-content: space-between; padding: 0 9px; border: 0; border-left: 2px solid transparent; border-radius: 2px; color: var(--muted); background: transparent; cursor: pointer; }
 .tree-children button::before { content: ""; position: absolute; top: 50%; left: -12px; width: 10px; height: 1px; background: #343740; }
 .tree-child-group { min-width: 0; }
-.tree-grandchild { height: 33px !important; padding-left: 20px !important; color: #777d87 !important; }
+.library-tree .tree-grandchild { height: 33px; padding-left: 20px; color: #777d87; }
 .tree-grandchild::after { content: ""; position: absolute; top: 50%; left: 9px; width: 4px; height: 4px; border-radius: 50%; background: #555a64; transform: translateY(-50%); }
 .library-tree button:hover { color: #d9dadd; background: rgba(255,255,255,.025); }
 .library-tree button.active { border-left-color: var(--red-bright); color: #f1f1f2; background: linear-gradient(90deg, rgba(212,59,52,.14), rgba(212,59,52,.02)); }
@@ -118,7 +119,7 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .workspace.sidebar-collapsed .sidebar-toggle { width: 40px; }
 .workspace.sidebar-collapsed .sidebar-toggle .icon { transform: none; }
 .workspace.sidebar-collapsed .source-heading { height: 48px; padding: 7px; }
-.workspace.sidebar-collapsed .source-select-wrap { height: 34px; margin: 0; }
+.workspace.sidebar-collapsed .source-select-wrap { height: 34px; }
 .workspace.sidebar-collapsed .source-select-wrap select { position: absolute; inset: 0; z-index: 2; padding: 0; opacity: 0; }
 .workspace.sidebar-collapsed .source-select-wrap > .icon:first-child { top: 8px; left: 13px; }
 .workspace.sidebar-collapsed .source-select-wrap > .icon:last-child { display: none; }
@@ -148,7 +149,7 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 
 .asset-panel { min-height: 0; display: flex; flex-direction: column; min-width: 0; border-right: 1px solid var(--line); background: #14161a; }
 .entity-panel, .media-panel { container-type: inline-size; }
-.asset-toolbar { height: 66px; display: flex; align-items: center; gap: 15px; padding: 0 18px; border-bottom: 1px solid var(--line-soft); }
+.asset-toolbar { height: 64px; display: flex; align-items: center; gap: 15px; padding: 0 18px; border-bottom: 1px solid var(--line-soft); }
 .search-box { height: 36px; flex: 1; display: flex; align-items: center; gap: 9px; padding: 0 10px; border: 1px solid var(--line); border-radius: 3px; color: var(--quiet); background: #101216; transition: border-color .15s; }
 .search-box:focus-within { border-color: #7f3b38; }
 .search-box input { min-width: 0; flex: 1; border: 0; outline: 0; color: #e6e7e9; background: none; font-size: 12px; }
@@ -160,7 +161,7 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .entity-search-input .search-box { width: 100%; }
 .entity-search-options { flex: 0 0 auto; display: flex; align-items: center; gap: 5px; }
 .entity-search-suggestions, .catalog-search-history { position: absolute; z-index: 35; top: calc(100% + 6px); left: 0; right: 0; max-height: min(430px, 58vh); padding: 5px; overflow-y: auto; border: 1px solid #42464f; border-radius: 4px; background: #17191e; box-shadow: 0 14px 34px rgba(0,0,0,.48); scrollbar-gutter: stable; }
-.entity-search-suggestions > button { width: 100%; min-height: 52px; display: grid; grid-template-columns: 42px minmax(100px, .7fr) minmax(130px, 1fr); align-items: center; gap: 9px; padding: 5px 9px 5px 6px; border: 1px solid transparent; border-left-width: 2px; border-radius: 3px; color: #b7bbc2; background: transparent; text-align: left; cursor: pointer; }
+.entity-search-suggestions > button { width: 100%; min-height: 52px; display: grid; grid-template-columns: 42px minmax(100px, 1.2fr) minmax(130px, 1fr); align-items: center; gap: 9px; padding: 5px 9px 5px 6px; border: 1px solid transparent; border-left-width: 2px; border-radius: 3px; color: #b7bbc2; background: transparent; text-align: left; cursor: pointer; }
 .entity-search-suggestions > button.catalog-suggestion-entity { border-left-color: var(--entity-accent); }
 .entity-search-suggestions > button.catalog-suggestion-voice { border-left-color: var(--voice-accent); }
 .entity-search-suggestions > button.catalog-suggestion-sound { border-left-color: var(--sound-accent); }
@@ -189,7 +190,7 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .catalog-history-entity { border-left-color: var(--entity-accent); }
 .catalog-history-voice { border-left-color: var(--voice-accent); }
 .catalog-history-sound { border-left-color: var(--sound-accent); }
-.catalog-history-target { min-width: 0; min-height: 48px; display: grid; grid-template-columns: 42px minmax(100px, .7fr) minmax(120px, 1fr); align-items: center; gap: 9px; padding: 5px 7px 5px 5px; border: 0; color: #b7bbc2; background: transparent; text-align: left; cursor: pointer; }
+.catalog-history-target { min-width: 0; min-height: 48px; display: grid; grid-template-columns: 42px minmax(100px, 1.2fr) minmax(120px, 1fr); align-items: center; gap: 9px; padding: 5px 7px 5px 5px; border: 0; color: #b7bbc2; background: transparent; text-align: left; cursor: pointer; }
 .catalog-history-target > em { min-width: 0; overflow: hidden; color: #8f789b; font-size: 12px; font-style: normal; text-align: right; text-overflow: ellipsis; white-space: nowrap; }
 .catalog-history-remove { width: 28px; height: 28px; align-self: center; display: grid; place-items: center; padding: 0; border: 1px solid transparent; border-radius: 3px; color: #656b75; background: transparent; cursor: pointer; }
 .catalog-history-remove:hover { border-color: #6b3a38; color: #e06a64; background: #281b1d; }
@@ -293,7 +294,7 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .asset-size { color: #777c85; font: 500 10px/1 var(--sans-condensed); text-align: right; }
 .asset-row > .icon { color: #4d5159; }
 .entity-row { grid-template-columns: 34px minmax(135px, 1fr) 70px 80px 12px; }
-.media-row { grid-template-columns: 34px minmax(160px, 1fr) minmax(110px, .45fr) 12px; }
+.media-row { grid-template-columns: 34px minmax(160px, 1.2fr) minmax(110px, 1fr) 12px; }
 .media-row.playing, .media-card.playing { background: linear-gradient(90deg, rgba(126,71,151,.16), rgba(255,255,255,.018)); }
 .media-links { min-width: 0; overflow: hidden; color: #888e98; font-size: 12px; text-overflow: ellipsis; white-space: nowrap; }
 .file-icon.format-audio { border-color: #4d3e58; color: #b58acb; background: #26212b; }
@@ -333,7 +334,7 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .no-results span { margin-top: 6px; font-size: 11px; }
 .no-results button { margin-top: 13px; border: 0; color: #c86d67; background: none; font-size: 11px; cursor: pointer; }
 
-.detail-panel { min-height: 0; padding: 12px 14px 14px; overflow-y: scroll; scrollbar-color: #555b66 #17191d; scrollbar-width: auto; scrollbar-gutter: stable; }
+.detail-panel { min-height: 0; padding: 12px 14px 14px; overflow-y: scroll; scrollbar-color: #555b66 #17191d; scrollbar-width: auto; scrollbar-gutter: stable; container-type: inline-size; }
 .detail-title { min-height: 32px; display: flex; align-items: center; justify-content: space-between; gap: 12px; }
 .detail-title > div { min-width: 0; }
 .detail-actions { flex: 0 0 auto; display: flex; align-items: center; gap: 7px; }
@@ -341,9 +342,20 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .detail-title .layout-toggle button { width: 28px; height: 28px; }
 .detail-heading-line { min-width: 0; display: flex; align-items: center; gap: 9px; overflow: hidden; }
 .audio-detail-title .detail-heading-line { flex: 1 1 auto; }
-.audio-header-player { width: 112px; min-width: 112px; height: 32px; display: grid; grid-template-columns: 30px minmax(0, 1fr) 30px; align-items: center; gap: 7px; }
-.audio-header-meta { min-width: 0; overflow: hidden; color: #898e97; font: 12px/1 var(--sans-condensed); text-align: center; text-overflow: ellipsis; white-space: nowrap; }
-.audio-header-player .audio-download-action { justify-self: end; }
+.audio-detail-player { display: flex; flex-wrap: wrap; align-items: center; gap: 9px; margin-top: 10px; padding: 7px 9px; border: 1px solid var(--line-soft); border-radius: 3px; background: #14161a; }
+.audio-detail-player > .compact-audio-player, .audio-detail-player > .audio-download-action { flex: 0 0 auto; }
+/* Mirrors .frame-controls input: a slim native slider, here filled up to the playhead. */
+.audio-progress { flex: 1 1 160px; min-width: 120px; height: 12px; margin: 0; appearance: none; background: transparent; cursor: pointer; }
+.audio-progress:disabled { cursor: default; }
+.audio-progress-time { flex: 0 0 auto; color: #9a9ea6; font: 500 12px/1 var(--sans-condensed); white-space: nowrap; }
+.audio-progress-time i { margin: 0 2px; color: #4f535b; font-style: normal; }
+.audio-progress::-webkit-slider-runnable-track { height: 3px; border-radius: 2px; background: linear-gradient(to right, var(--voice-accent) var(--audio-progress, 0%), var(--line) var(--audio-progress, 0%)); }
+.audio-progress::-webkit-slider-thumb { width: 9px; height: 9px; margin-top: -3px; border: 0; border-radius: 50%; background: var(--voice-accent); appearance: none; }
+.audio-progress::-moz-range-track { height: 3px; border-radius: 2px; background: var(--line); }
+.audio-progress::-moz-range-progress { height: 3px; border-radius: 2px; background: var(--voice-accent); }
+.audio-progress::-moz-range-thumb { width: 9px; height: 9px; border: 0; border-radius: 50%; background: var(--voice-accent); }
+.audio-progress:hover::-webkit-slider-thumb { background: #e3c7ef; }
+.audio-progress:hover::-moz-range-thumb { background: #e3c7ef; }
 .format-pill { display: inline-block; padding: 3px 6px; border: 1px solid #61312f; border-radius: 2px; color: #d77772; background: rgba(187,52,47,.08); font: 600 8px/1 var(--sans-condensed); letter-spacing: .11em; text-transform: uppercase; }
 .detail-title h2 { min-width: 0; margin: 0; overflow: hidden; color: #f0f0f1; font: 600 18px/1.15 var(--sans-condensed); letter-spacing: .015em; text-overflow: ellipsis; white-space: nowrap; }
 .detail-title > div > small { flex: 0 1 auto; overflow: hidden; color: #747983; font-size: 12px; text-overflow: ellipsis; white-space: nowrap; }
@@ -395,7 +407,7 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .text-preview { margin-top: 19px; border: 1px solid var(--line); background: #101216; }
 .text-preview label { height: 34px; display: flex; align-items: center; gap: 8px; padding: 0 10px; border-bottom: 1px solid var(--line); color: var(--quiet); }
 .text-preview input { min-width: 0; flex: 1; border: 0; outline: 0; color: #d7d9dc; background: transparent; font-size: 10px; }
-.text-preview pre { height: 250px; margin: 0; padding: 11px; overflow: scroll; color: #aeb4aa; font: 12px/1.55 "Cascadia Mono", Consolas, monospace; white-space: pre-wrap; word-break: break-word; scrollbar-color: #555b66 #17191d; scrollbar-gutter: stable; }
+.text-preview pre { height: 250px; margin: 0; padding: 11px; overflow: scroll; color: #aeb4aa; font: 12px/1.55 var(--mono); white-space: pre-wrap; word-break: break-word; scrollbar-color: #555b66 #17191d; scrollbar-gutter: stable; }
 .text-preview > small { display: block; padding: 7px 10px; border-top: 1px solid var(--line-soft); color: #666b74; font-size: 8px; }
 .metadata { margin-top: 20px; }
 .metadata h3 { margin: 0 0 8px; color: #aeb1b7; font: 600 10px/1 var(--sans-condensed); letter-spacing: .14em; text-transform: uppercase; }
@@ -441,7 +453,7 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
   .sound-transcript-tag dd { white-space: normal; overflow-wrap: anywhere; }
 }
 .detail-panel-wide .preview-block, .detail-panel-wide .video-preview, .detail-panel-wide .text-preview { max-width: 860px; margin-right: auto; margin-left: auto; }
-.mono { font-family: "Cascadia Mono", Consolas, monospace; letter-spacing: .05em; }
+.mono { font-family: var(--mono); letter-spacing: .05em; }
 .export-button { width: 100%; margin-top: 13px; background: #1e2025; }
 .unsupported-preview { height: 190px; margin-top: 19px; display: flex; flex-direction: column; align-items: center; justify-content: center; border: 1px dashed #363940; color: #646972; background: #14161a; }
 .unsupported-preview strong { margin-top: 12px; color: #aeb1b7; font-size: 12px; }
@@ -498,7 +510,7 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .entity-preview-controls { position: absolute; z-index: 5; left: 50%; bottom: 10px; width: min(calc(100% - 20px), 620px); display: flex; align-items: stretch; justify-content: center; gap: 6px; transform: translateX(-50%); }
 .entity-preview-controls > .frame-controls { min-width: 0; flex: 1 1 auto; border-top: 1px solid var(--line); border-radius: 3px; background: rgba(22,24,28,.91); box-shadow: 0 5px 18px rgba(0,0,0,.28); backdrop-filter: blur(5px); }
 .animation-candidate-control { min-width: 0; min-height: 43px; flex: 0 1 190px; display: grid; grid-template-columns: auto minmax(72px, 1fr); align-items: center; gap: 7px; padding: 6px 8px; border: 1px solid var(--line); border-radius: 3px; color: #858a94; background: rgba(22,24,28,.94); box-shadow: 0 5px 18px rgba(0,0,0,.28); font-size: 12px; backdrop-filter: blur(5px); }
-.animation-candidate-control select { min-width: 0; width: 100%; height: 29px; padding: 0 6px; overflow: hidden; border: 1px solid #3b3f47; border-radius: 2px; outline: 0; color: #d3d5d9; background: #111317; font: 12px/1 "Cascadia Mono", Consolas, monospace; text-overflow: ellipsis; }
+.animation-candidate-control select { min-width: 0; width: 100%; height: 29px; padding: 0 6px; overflow: hidden; border: 1px solid #3b3f47; border-radius: 2px; outline: 0; color: #d3d5d9; background: #111317; font: 12px/1 var(--mono); text-overflow: ellipsis; }
 .entity-detail-narrow .animation-candidate-control { flex-basis: 108px; grid-template-columns: minmax(0, 1fr); }
 .entity-detail-narrow .animation-candidate-control > span { display: none; }
 .entity-effect-overlay { position: absolute; z-index: 3; inset: 0; overflow: hidden; pointer-events: none; transition: opacity 40ms linear; }
@@ -600,7 +612,7 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .format-checks .file-icon { width: 27px; height: 27px; }
 .format-checks strong { min-width: 0; overflow: hidden; font-size: 12px; font-weight: 600; text-overflow: ellipsis; white-space: nowrap; }
 .format-checks em { color: #7c818a; font-size: 11px; font-style: normal; }
-.settings-empty { min-height: 52px; display: grid; place-items: center; padding: 12px; border: 1px dashed #343840; color: #777d86; background: #15171a; font-size: 13px; text-align: center; }
+.settings-empty { margin-top: 10px; min-height: 52px; display: grid; place-items: center; padding: 12px; border: 1px dashed #343840; color: #777d86; background: #15171a; font-size: 13px; text-align: center; }
 .settings-source-list, .resource-pack-list { display: grid; gap: 7px; }
 .settings-source-list article, .resource-pack-list article { min-width: 0; min-height: 52px; display: grid; grid-template-columns: 31px minmax(0, 1fr) auto; align-items: center; gap: 10px; padding: 7px 9px; border: 1px solid var(--line-soft); background: #15171b; }
 .settings-source-list article.active { border-color: #513735; background: #1d191c; }
@@ -609,7 +621,7 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .settings-source-list strong, .resource-pack-list strong { color: #d5d7db; font-size: 13px; }
 .settings-source-list small, .resource-pack-list small { margin-top: 4px; color: #777d86; font-size: 12px; }
 .settings-source-list article > button { min-height: 31px; padding: 0 9px; border: 1px solid var(--line); border-radius: 3px; color: #aaaeb5; background: #1c1e23; font-size: 12px; cursor: pointer; }
-.settings-source-list article > button:disabled { opacity: .55; cursor: wait; }
+.settings-source-list article > button:disabled { opacity: .55; cursor: not-allowed; }
 .settings-discoveries { display: grid; grid-template-columns: repeat(2, minmax(0, 1fr)); gap: 7px; margin-top: 12px; }
 .settings-discoveries button { min-width: 0; min-height: 52px; display: grid; grid-template-columns: 25px minmax(0, 1fr) auto; align-items: center; gap: 8px; padding: 7px 9px; border: 1px solid #344039; color: #82a98d; background: #151a17; text-align: left; cursor: pointer; }
 .settings-discoveries button > span { min-width: 0; }
@@ -670,6 +682,8 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .asset-card-preview { position: relative; width: 100%; height: 120px; flex: 0 0 auto; display: grid; place-items: center; overflow: hidden; color: #737983; background-color: #111317; background-image: linear-gradient(45deg,#181a1f 25%,transparent 25%),linear-gradient(-45deg,#181a1f 25%,transparent 25%),linear-gradient(45deg,transparent 75%,#181a1f 75%),linear-gradient(-45deg,transparent 75%,#181a1f 75%); background-size: 16px 16px; background-position: 0 0,0 8px,8px -8px,-8px 0; }
 .asset-card-preview > img { position: absolute; inset: 0; display: block; width: 100%; height: 100%; padding: 10px; image-rendering: pixelated; object-fit: contain; filter: drop-shadow(0 6px 10px rgba(0,0,0,.52)); }
 .asset-card-preview.format-bag_audio, .asset-card-preview.format-wav, .asset-card-preview.format-aud { color: #b78acb; background: linear-gradient(145deg,#211b27,#14161a); }
+/* Movie frames are photographic, not pixel art. */
+.asset-card-preview.format-video > img { image-rendering: auto; }
 .asset-card-copy { min-width: 0; padding: 10px 11px 11px; }
 .asset-card-copy strong, .asset-card-copy small { display: block; overflow: hidden; text-overflow: ellipsis; white-space: nowrap; }
 .asset-card-copy strong { color: #dddfe2; font-size: 13px; font-weight: 600; }
@@ -681,7 +695,7 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .load-more { width: calc(100% - 24px); min-height: 38px; margin: 12px; border: 1px solid var(--line); border-radius: 3px; color: #aeb2b9; background: #1b1d22; font-size: 12px; cursor: pointer; }
 .asset-grid .load-more { width: 100%; margin: 0; }
 .load-more:hover:not(:disabled) { border-color: #555a64; color: #eef0f2; }
-.load-more:disabled { opacity: .6; cursor: wait; }
+.load-more:disabled { opacity: .6; cursor: not-allowed; }
 .entity-card-preview { height: 135px; }
 .entity-card-preview.catalog-suggestion-thumbnail { width: 38px; height: 38px; border: 1px solid #47779b; border-radius: 4px; background-size: 8px 8px; }
 .entity-card-preview.catalog-suggestion-thumbnail > img { padding: 3px; filter: drop-shadow(0 2px 3px rgba(0,0,0,.5)); }
@@ -704,8 +718,7 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .voxel-canvas, .voxel-canvas canvas { width: 100%; height: 100%; display: block; }
 .voxel-canvas canvas { cursor: grab; }
 .voxel-canvas canvas:active { cursor: grabbing; }
-.voxel-reset { position: absolute; z-index: 2; top: 10px; right: 10px; min-height: 29px; display: inline-flex; align-items: center; padding: 0 9px; border: 1px solid rgba(91,96,107,.75); border-radius: 3px; color: #c4c7cd; background: rgba(15,17,21,.82); font-size: 11px; backdrop-filter: blur(4px); }
-.voxel-reset { cursor: pointer; }
+.voxel-reset { position: absolute; z-index: 2; top: 10px; right: 10px; min-height: 29px; display: inline-flex; align-items: center; padding: 0 9px; border: 1px solid rgba(91,96,107,.75); border-radius: 3px; color: #c4c7cd; background: rgba(15,17,21,.82); font-size: 11px; cursor: pointer; backdrop-filter: blur(4px); }
 .voxel-reset:hover { border-color: #8d4a46; color: #f0f1f2; }
 .voxel-status { position: absolute; inset: 0; display: grid; place-items: center; color: #a8acb4; background: rgba(15,17,21,.72); font-size: 12px; }
 .voxel-status.error { color: #df7a74; }
@@ -713,7 +726,7 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 
 .entity-tags { display: flex; flex-wrap: wrap; gap: 7px; padding: 0 9px 10px; }
 .entity-tags > span { min-height: 28px; display: inline-flex; align-items: center; gap: 5px; padding: 0 9px; border: 1px solid #383b43; border-radius: 14px; color: #aeb2ba; background: #16181d; font-size: 12px; }
-.entity-tags code { color: #dc7771; font: 600 11px/1 "Cascadia Mono", Consolas, monospace; }
+.entity-tags code { color: #dc7771; font: 600 11px/1 var(--mono); }
 .entity-section { margin-top: 12px; border: 1px solid var(--line-soft); background: #14161a; }
 .entity-section > summary, .dependency-groups > details > summary { min-height: 40px; display: flex; align-items: center; gap: 9px; padding: 0 11px; color: #c8cbd0; cursor: pointer; list-style: none; }
 .entity-section > summary::-webkit-details-marker, .dependency-groups summary::-webkit-details-marker { display: none; }
@@ -727,17 +740,22 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .media-association-list header strong { flex: 0 0 auto; color: #b38ac5; font-size: 12px; font-weight: 600; }
 .media-association-slots { min-width: 0; display: flex; align-items: center; gap: 4px; overflow: hidden; }
 .media-association-slots strong { padding: 3px 6px; border: 1px solid #403746; border-radius: 10px; background: #211c25; white-space: nowrap; }
-.media-association-list header code { overflow: hidden; color: #cfd1d5; font: 600 12px/1 "Cascadia Mono", Consolas, monospace; text-overflow: ellipsis; white-space: nowrap; }
+.media-association-list header code { overflow: hidden; color: #cfd1d5; font: 600 12px/1 var(--mono); text-overflow: ellipsis; white-space: nowrap; }
 .media-association-list header em { margin-left: auto; color: #727781; font-size: 11px; font-style: normal; }
-.media-sample { min-width: 0; min-height: 40px; display: grid; grid-template-columns: minmax(90px, .55fr) 30px minmax(0, 2.4fr) 30px; align-items: center; gap: 10px; padding: 5px 8px; }
+.media-sample { min-width: 0; min-height: 40px; display: grid; grid-template-columns: minmax(90px, 1fr) auto minmax(160px, 2fr) 30px; align-items: center; gap: 10px; padding: 5px 8px; }
+.media-sample-no-audio { grid-template-columns: minmax(90px, 1fr) minmax(0, 2fr); }
 .media-sample + .media-sample { border-top: 1px solid var(--line-soft); }
-.media-sample-id { min-width: 0; overflow: hidden; color: #d3d5d9; font: 600 12px/1 "Cascadia Mono", Consolas, monospace; text-overflow: ellipsis; white-space: nowrap; }
+.media-sample-id { min-width: 0; overflow: hidden; color: #d3d5d9; font: 600 12px/1 var(--mono); text-overflow: ellipsis; white-space: nowrap; }
 .media-sample-id em { margin-left: 5px; color: #8f7598; font: 500 11px/1 var(--sans-condensed); font-style: normal; }
 .media-sample-texts { min-width: 0; display: grid; grid-template-columns: minmax(0, 1fr); gap: 8px; }
 .media-sample-texts.bilingual { grid-template-columns: minmax(0, 1fr); gap: 3px; }
 .media-sample-copy { min-width: 0; overflow: hidden; color: #c7c9cd; font-size: 12px; line-height: 1.35; text-overflow: ellipsis; white-space: nowrap; }
 .media-sample-copy b { margin-right: 6px; color: #737985; font-size: 11px; font-weight: 500; }
-.media-sample-missing { color: #8f755a; font-size: 11px; font-style: normal; }
+.media-sample-id .media-sample-missing { color: #8f755a; }
+
+@container (max-width: 640px) {
+  .media-sample-texts.bilingual { grid-template-columns: minmax(0, 1fr); }
+}
 .compact-audio-player { width: 30px; height: 30px; display: inline-grid; place-items: center; }
 .compact-audio-player audio { display: none; }
 .compact-audio-player button, .audio-download-action { width: 30px; height: 30px; display: grid; place-items: center; padding: 0; border: 1px solid #3d3a43; border-radius: 3px; color: #b68ac9; background: #211d25; text-decoration: none; cursor: pointer; }
@@ -761,13 +779,13 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .animation-thumbnail.body-action { color: #84a98c; background: #18201b; }
 .dependency-groups { display: grid; gap: 7px; }
 .dependency-groups > details { border: 1px solid var(--line-soft); background: #14171a; }
-.dependency-groups summary strong { min-width: 0; overflow: hidden; color: #d5d7da; font: 600 12px/1 "Cascadia Mono", Consolas, monospace; text-overflow: ellipsis; white-space: nowrap; }
+.dependency-groups summary strong { min-width: 0; overflow: hidden; color: #d5d7da; font: 600 12px/1 var(--mono); text-overflow: ellipsis; white-space: nowrap; }
 .dependency-compact { display: grid; gap: 6px; padding: 0 8px 8px; }
 .dependency-compact article { padding: 8px; border: 1px solid #2b2e34; background: #181a1e; }
 .dependency-compact article.unresolved { border-color: #6e5130; }
 .dependency-compact header { display: flex; align-items: center; gap: 8px; }
 .dependency-compact header span { color: #858b94; font-size: 12px; }
-.dependency-compact header code { color: #d2d5d9; font: 600 12px/1 "Cascadia Mono", Consolas, monospace; }
+.dependency-compact header code { color: #d2d5d9; font: 600 12px/1 var(--mono); }
 .dependency-compact header em { margin-left: auto; color: var(--amber); font-size: 11px; font-style: normal; }
 .property-tags { display: flex; flex-wrap: wrap; gap: 5px; margin-top: 8px; }
 .property-tags span { max-width: 100%; display: inline-flex; gap: 5px; padding: 5px 7px; overflow: hidden; border: 1px solid #30343a; border-radius: 12px; color: #aeb2b9; font-size: 11px; text-overflow: ellipsis; white-space: nowrap; }
@@ -818,12 +836,12 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .detached-shell > .entity-detail-wide { overflow-y: hidden; }
 .detached-error { padding: 28px; color: #df7a74; font-size: 14px; }
 .asset-associations { margin-top: 18px; }
-.asset-associations h3 { margin: 0 0 8px; color: #aeb1b7; font: 600 11px/1 var(--sans-condensed); letter-spacing: .12em; text-transform: uppercase; }
+.asset-associations h3 { margin: 0 0 8px; color: #aeb1b7; font: 600 11px/1 var(--sans-condensed); letter-spacing: .14em; text-transform: uppercase; }
 .asset-associations > div { display: flex; flex-wrap: wrap; gap: 7px; }
 .asset-associations article { min-width: 0; max-width: 100%; padding: 8px 10px; border: 1px solid #39343e; border-radius: 3px; background: #19171d; }
 .asset-associations article span { display: block; color: #ad83c0; font-size: 11px; }
 .asset-associations article strong, .asset-associations article code { display: block; margin-top: 4px; overflow: hidden; color: #d4d6da; font-size: 12px; text-overflow: ellipsis; white-space: nowrap; }
-.asset-associations article code { color: #8f949d; font-family: "Cascadia Mono", Consolas, monospace; }
+.asset-associations article code { color: #8f949d; font-family: var(--mono); }
 .asset-associations article p { margin: 6px 0 0; color: #e2e3e5; font-size: 13px; line-height: 1.45; }
 .asset-associations article p b { margin-right: 7px; color: #8e749b; font-size: 11px; font-weight: 500; }
 .detail-panel-wide .asset-associations:not(.audio-associations) > div { display: grid; grid-template-columns: repeat(auto-fill, minmax(230px, 1fr)); }
@@ -836,8 +854,13 @@ button:focus-visible, input:focus-visible, select:focus-visible, a:focus-visible
 .audio-association-texts > div { width: 100%; min-width: 0; padding: 5px 0; display: grid; grid-template-columns: max-content minmax(0, 1fr); align-items: start; gap: 10px; }
 .audio-association-texts dt { color: #94749f; font-size: 12px; line-height: 1.5; }
 .audio-association-texts dd { min-width: 0; margin: 0; color: #d8d9dc; font-size: 13px; line-height: 1.5; white-space: pre-wrap; overflow-wrap: anywhere; }
-.video-preview { min-height: 230px; margin-top: 19px; display: grid; place-items: center; overflow: hidden; border: 1px solid var(--line); background: #0d0f12; }
+.video-preview { position: relative; min-height: 230px; margin-top: 19px; display: grid; place-items: center; overflow: hidden; border: 1px solid var(--line); background: #0d0f12; }
 .video-preview video { width: 100%; max-height: 360px; display: block; }
+.video-poster { width: 100%; max-height: 360px; display: block; object-fit: contain; }
+/* Floating overlay control, following .voxel-reset; kept free of .button so
+   its centering transform is never overridden by .button:active. */
+.video-convert { position: absolute; z-index: 1; left: 50%; top: 50%; width: 54px; height: 54px; display: grid; place-items: center; padding: 0; border: 1px solid rgba(140,145,155,.55); border-radius: 50%; color: #e9ebee; background: rgba(13,15,18,.66); cursor: pointer; backdrop-filter: blur(3px); transform: translate(-50%, -50%); transition: border-color .15s, background .15s, color .15s; }
+.video-convert:hover { border-color: #8d4a46; color: #f0f1f2; background: rgba(15,17,21,.88); }
 .video-error { color: #d97570; font-size: 13px; }
 
 .asset-list::-webkit-scrollbar, .detail-panel::-webkit-scrollbar, .entity-detail-sections::-webkit-scrollbar, .library-tree::-webkit-scrollbar, .tag-filter::-webkit-scrollbar, .text-preview pre::-webkit-scrollbar, .search-results-scroll::-webkit-scrollbar, .entity-search-suggestions::-webkit-scrollbar, .catalog-search-history::-webkit-scrollbar { width: 11px; height: 7px; }
@@ -921,11 +944,11 @@ body { font-size: 14px; }
   .source-panel { display: block; border-right: 0; }
   .sidebar-brand { min-height: 58px; }
   .sidebar-settings { margin-bottom: 9px; }
-  .source-heading { padding: 14px 13px 11px; }
+  .source-heading { padding: 11px 13px; }
   .summary-grid, .reference-card { display: none; }
   .library-tree { max-height: 310px; }
   .asset-panel { height: 480px; }
-  .asset-toolbar { height: 57px; padding: 0 12px; }
+  .asset-toolbar { height: 58px; padding: 0 12px; }
   .entity-filter { width: 86px; }
   .result-count { display: none; }
   .asset-row { padding-left: 12px; }
@@ -935,8 +958,8 @@ body { font-size: 14px; }
   .preview-stage { height: 300px; }
   .media-group-items.list-columns { grid-template-columns: 1fr; }
   .media-group-items.list-columns > .asset-row:nth-child(odd) { border-right: 0; }
-  .media-sample { grid-template-columns: minmax(90px, 1fr) 30px 30px; }
-  .media-sample .compact-audio-player, .media-sample .media-sample-missing { grid-column: 2; }
+  .media-sample { grid-template-columns: minmax(90px, 1fr) auto 30px; }
+  .media-sample .compact-audio-player { grid-column: 2; }
   .media-sample .audio-download-action { grid-column: 3; }
   .media-sample-texts { grid-column: 1 / -1; grid-row: 2; }
   .media-sample-texts.bilingual { grid-template-columns: minmax(0, 1fr); }
diff --git a/src/ra2_explorer/api.py b/src/ra2_explorer/api.py
index 0705632..da7c322 100644
--- a/src/ra2_explorer/api.py
+++ b/src/ra2_explorer/api.py
@@ -6,6 +6,7 @@
 import os
 import re
 import uuid
+from collections.abc import Collection
 from pathlib import Path
 from typing import Literal
 from urllib.parse import quote
@@ -53,6 +54,7 @@
     BUNDLED_UNIT_VOICE_TRANSCRIPT_PATH,
     BUNDLED_VOICE_TRANSLATION_PATH,
     load_audio_transcript,
+    load_audio_translations,
     load_known_names,
     reference_status,
     sync_known_names,
@@ -126,6 +128,12 @@ def __init__(self, settings: Settings):
                     BUNDLED_VOICE_TRANSLATION_PATH,
                 ),
             ),
+            load_audio_translations(
+                supplement_paths=(
+                    settings.mission_audio_transcript_path,
+                    BUNDLED_VOICE_TRANSLATION_PATH,
+                )
+            ),
         )
         self.video = VideoTranscoder(self.database, self.reader, self.derived)
 
@@ -381,7 +389,7 @@ def media(
         limit: int = Query(default=500, ge=1, le=500),
         offset: int = Query(default=0, ge=0),
     ) -> dict[str, object]:
-        if kind is not None and kind not in {"voice", "sound", "unknown"}:
+        if kind is not None and kind not in {"voice", "sound", "music", "unknown"}:
             raise HTTPException(status_code=422, detail="未知音频类型")
         if sort not in {"name_asc", "name_desc", "description_asc"}:
             raise HTTPException(status_code=422, detail="未知音频排序方式")
@@ -770,6 +778,7 @@ def asset_preview(
             "tmp",
             "pcx",
             "map",
+            "vpl",
         }:
             raise HTTPException(status_code=409, detail="该格式没有图像预览")
         artifact_path = _asset_artifact_path(
@@ -795,6 +804,8 @@ def asset_preview(
         asset_record, data = services.reader.read(asset_id)
         if asset_record["format"] == "pal":
             image = parse_palette(data).preview(cell_size=max(4, scale * 3))
+        elif asset_record["format"] == "vpl":
+            image = parse_vpl(data).preview(cell_size=max(1, scale))
         elif asset_record["format"] == "shp":
             sprite = parse_shp(data)
             if frame >= len(sprite.frames) or (
@@ -806,6 +817,7 @@ def asset_preview(
                 asset_record,
                 palette_id,
                 palette_kind,
+                used_indices=set(sprite.pixels(frame)),
             )
             if player_color:
                 palette = (palette or grayscale_palette()).with_player_color(player_color)
@@ -901,6 +913,15 @@ def asset_video(asset_id: str) -> FileResponse:
             headers={"Cache-Control": "private, max-age=86400"},
         )
 
+    @app.get("/api/assets/{asset_id}/poster.png")
+    def asset_poster(asset_id: str) -> FileResponse:
+        path = services.video.poster_frame(asset_id)
+        return FileResponse(
+            path,
+            media_type="image/png",
+            headers={"Cache-Control": "private, max-age=86400"},
+        )
+
     @app.get("/api/palettes")
     def palettes(source_id: str) -> list[dict[str, object]]:
         return services.database.palette_assets(source_id)
@@ -1199,6 +1220,7 @@ def _select_palette(
     asset: dict[str, object],
     palette_id: str | None,
     palette_kind: Literal["unit", "animation"] | None = None,
+    used_indices: Collection[int] | None = None,
 ):
     palettes = services.database.palette_assets(str(asset["source_id"]))
     if palette_id:
@@ -1245,17 +1267,45 @@ def _select_palette(
         else [f"unit{theater}.pal", "unittem.pal", f"iso{theater}.pal", "isotem.pal"]
     )
     priority = {name: index for index, name in enumerate(preferred)}
-    palette_asset = min(
-        palettes,
+    candidates = sorted(
+        (item for item in palettes if str(item["display_name"]).lower() in priority),
         key=lambda item: (
-            priority.get(str(item["display_name"]).lower(), 99),
-            item["display_name"],
+            priority[str(item["display_name"]).lower()],
+            str(item["display_name"]),
         ),
     )
+    if not candidates:
+        candidates = sorted(palettes, key=lambda item: str(item["display_name"]))
+    # Unit palettes mark indices 204-239 as magenta placeholders; terrain and
+    # interface art legitimately paints there. With no explicit intent, prefer
+    # the candidate whose placeholders the artwork leaves unused.
+    if palette_kind is None and used_indices and len(candidates) > 1:
+        candidates = _least_placeholder_palettes(services, candidates, used_indices)
+    palette_asset = candidates[0]
     _, palette_data = services.reader.read(palette_asset["id"])
     return parse_palette(palette_data)
 
 
+def _least_placeholder_palettes(
+    services: Services,
+    candidates: list[dict[str, object]],
+    used_indices: Collection[int],
+) -> list[dict[str, object]]:
+    best_hits = None
+    for candidate in candidates:
+        _, palette_data = services.reader.read(str(candidate["id"]))
+        placeholders = parse_palette(palette_data).placeholder_indices()
+        hits = sum(1 for index in used_indices if index in placeholders)
+        if best_hits is None or hits < best_hits:
+            best_hits = hits
+            best = [candidate]
+        elif hits == best_hits:
+            best.append(candidate)
+        if not best_hits:
+            break
+    return best
+
+
 def _alpha_composite_centered(layers: list[Image.Image]) -> Image.Image:
     width = max(layer.width for layer in layers)
     height = max(layer.height for layer in layers)
diff --git a/src/ra2_explorer/codecs/pal.py b/src/ra2_explorer/codecs/pal.py
index c90b7d7..c48faf4 100644
--- a/src/ra2_explorer/codecs/pal.py
+++ b/src/ra2_explorer/codecs/pal.py
@@ -9,6 +9,9 @@
 
 PALETTE_COLORS = 256
 PALETTE_BYTES = PALETTE_COLORS * 3
+# RA2 unit palettes fill indices 204-239 with this marker instead of real art,
+# so art painted with those indices needs a terrain palette to render.
+PLACEHOLDER_MAGENTA = (252, 0, 252)
 PLAYER_COLOR_PRESETS = {
     "red": (214, 59, 52),
     "blue": (61, 111, 198),
@@ -44,6 +47,14 @@ def preview(self, *, cell_size: int = 14) -> Image.Image:
                     pixels[x, y] = color
         return image
 
+    def placeholder_indices(self) -> frozenset[int]:
+        """Indices filled with the magenta marker, which are never real art."""
+        return frozenset(
+            index
+            for index in range(1, PALETTE_COLORS)
+            if self.colors[index] == PLACEHOLDER_MAGENTA
+        )
+
     def remap(
         self,
         color: tuple[int, int, int],
@@ -114,6 +125,7 @@ def grayscale_palette() -> Palette:
 
 
 __all__ = [
+    "PLACEHOLDER_MAGENTA",
     "PLAYER_COLOR_PRESETS",
     "Palette",
     "grayscale_palette",
diff --git a/src/ra2_explorer/codecs/sniff.py b/src/ra2_explorer/codecs/sniff.py
index 579c407..676afda 100644
--- a/src/ra2_explorer/codecs/sniff.py
+++ b/src/ra2_explorer/codecs/sniff.py
@@ -63,6 +63,8 @@ def sniff_format(data: bytes | bytearray | memoryview, name: str | None = None)
     if looks_like_shp(view):
         return "shp"
     sample = bytes(view[:4096])
+    if sample.startswith(b"BIK"):
+        return "video"
     if sample.startswith(b"RIFF") and sample[8:12] == b"WAVE":
         return "wav"
     if sample.startswith(b"GABA"):
diff --git a/src/ra2_explorer/codecs/vpl.py b/src/ra2_explorer/codecs/vpl.py
index 05bbc90..b262837 100644
--- a/src/ra2_explorer/codecs/vpl.py
+++ b/src/ra2_explorer/codecs/vpl.py
@@ -2,6 +2,8 @@
 
 from dataclasses import dataclass
 
+from PIL import Image
+
 from ra2_explorer.codecs.binary import BinaryReader, checked_product
 from ra2_explorer.codecs.pal import Palette, parse_palette
 from ra2_explorer.errors import InvalidFormatError
@@ -28,6 +30,27 @@ def color_index(self, section: int, color: int) -> int:
         selected = max(0, min(len(self.sections) - 1, section))
         return self.sections[selected][color & 0xFF]
 
+    def preview(self, *, cell_size: int = 4) -> Image.Image:
+        """Render the lighting lookup as one row per section."""
+        if cell_size < 1:
+            raise ValueError("cell_size must be positive")
+        rows = max(1, self.section_count)
+        image = Image.new(
+            "RGB", (LOOKUP_WIDTH * cell_size, rows * cell_size)
+        )
+        pixels = image.load()
+        for row in range(rows):
+            for color in range(LOOKUP_WIDTH):
+                value = self.color_index(row, color)
+                shade = (value, value, value)
+                for offset_y in range(cell_size):
+                    for offset_x in range(cell_size):
+                        pixels[
+                            color * cell_size + offset_x,
+                            row * cell_size + offset_y,
+                        ] = shade
+        return image
+
 
 def parse_vpl(data: bytes | bytearray | memoryview) -> VplFile:
     reader = BinaryReader(data, format_name="VPL")
diff --git a/src/ra2_explorer/library.py b/src/ra2_explorer/library.py
index dec3e82..a7f4ff0 100644
--- a/src/ra2_explorer/library.py
+++ b/src/ra2_explorer/library.py
@@ -10,6 +10,7 @@
 from ra2_explorer.codecs.bag import parse_bag_index
 from ra2_explorer.codecs.mix import (
     MixHashType,
+    MixIndex,
     classic_mix_hash,
     parse_local_mix_database,
     parse_mix,
@@ -35,6 +36,24 @@
 MAX_NESTING_DEPTH = 6
 MAX_ARCHIVES_PER_SOURCE = 4096
 MAX_SNIFF_BYTES = 67_108_864
+# Smallest plausible archive: 6-byte header plus one 12-byte entry and payload.
+MIN_MIX_PROBE_SIZE = 64
+# Audio bag payloads carry no magic bytes and can accidentally satisfy MIX
+# header validation, so they are recognized by CRC pairing with AUDIO.IDX.
+AUDIO_INDEX_CRCES = frozenset(
+    crc for name in ("audio.idx",) for crc in (ra2_mix_hash(name), classic_mix_hash(name))
+)
+AUDIO_BAG_CRCES = frozenset(
+    crc for name in ("audio.bag",) for crc in (ra2_mix_hash(name), classic_mix_hash(name))
+)
+
+
+def _probe_mix(payload: memoryview) -> MixIndex | None:
+    """Return a parsed MIX index for payloads that parse as MIX, else None."""
+    try:
+        return parse_mix(payload)
+    except Ra2ExplorerError:
+        return None
 
 
 class NameResolver:
@@ -302,12 +321,27 @@ def _index_archive(
             )
         )
 
-        direct_assets: dict[str, tuple[AssetRecord, memoryview]] = {}
+        audio_index_pair: tuple[AssetRecord, memoryview] | None = None
+        audio_bag_pair: tuple[AssetRecord, memoryview] | None = None
         for entry in index.entries:
             name = resolved_names.get(entry.crc)
             payload = index.payload(data, entry)
             asset_format = format_from_name(name)
             confidence = "name" if asset_format else "unknown"
+            is_audio_bag = entry.crc in AUDIO_BAG_CRCES
+            if (
+                name is None
+                and not is_audio_bag
+                and entry.size >= MIN_MIX_PROBE_SIZE
+                and _probe_mix(payload) is not None
+            ):
+                # Unnamed retail entries inside ra2.mix/ra2md.mix/language.mix
+                # are often encrypted nested MIX archives. Their headers carry
+                # no recognizable extension, so probe the MIX parser before
+                # falling back to content sniffing; otherwise the archive is
+                # misclassified and its members are lost to the index.
+                asset_format = "mix"
+                confidence = "content"
             if not asset_format and entry.size <= MAX_SNIFF_BYTES:
                 asset_format = sniff_format(payload, name)
                 if asset_format != "binary":
@@ -336,8 +370,10 @@ def _index_archive(
                 loose_relative_path=None,
             )
             result.assets.append(asset_record)
-            if name:
-                direct_assets[Path(name).name.casefold()] = (asset_record, payload)
+            if entry.crc in AUDIO_INDEX_CRCES:
+                audio_index_pair = (asset_record, payload)
+            elif is_audio_bag:
+                audio_bag_pair = (asset_record, payload)
 
             if asset_format != "mix" or depth >= MAX_NESTING_DEPTH:
                 continue
@@ -353,15 +389,13 @@ def _index_archive(
                 result=result,
             )
 
-        audio_index = direct_assets.get("audio.idx")
-        audio_bag = direct_assets.get("audio.bag")
-        if audio_index and audio_bag:
+        if audio_index_pair is not None and audio_bag_pair is not None:
             try:
                 self._expand_audio_bag(
                     source_id,
-                    audio_bag[0],
-                    audio_index[1],
-                    len(audio_bag[1]),
+                    audio_bag_pair[0],
+                    audio_index_pair[1],
+                    audio_bag_pair[0].size,
                     result,
                 )
             except Ra2ExplorerError as error:
diff --git a/src/ra2_explorer/reference_data.py b/src/ra2_explorer/reference_data.py
index cabd283..f9ae2a7 100644
--- a/src/ra2_explorer/reference_data.py
+++ b/src/ra2_explorer/reference_data.py
@@ -51,16 +51,26 @@
 
 BUILTIN_NAMES = (
     "ra2.mix",
+    "ra2md.mix",
     "language.mix",
+    "langmd.mix",
     "multi.mix",
     "cache.mix",
+    "cachemd.mix",
     "local.mix",
     "neutral.mix",
     "conquer.mix",
     "generic.mix",
     "isogen.mix",
+    "sidec01.mix",
+    "sidec02.mix",
+    "audiomd.mix",
     "cameo.mix",
     "audio.mix",
+    "snow.mix",
+    "urban.mix",
+    "desert.mix",
+    "lunar.mix",
     "rules.ini",
     "art.ini",
     "ra2.csf",
@@ -170,7 +180,7 @@ def load_audio_transcript(
         if original_text and not translated_text:
             translated_text = translations_by_original.get(original_text.strip()) or (
                 translations_by_normalized_original.get(
-                    _audio_translation_lookup_key(original_text)
+                    audio_translation_lookup_key(original_text)
                 )
             )
             if translated_text:
@@ -182,6 +192,31 @@ def load_audio_transcript(
     return entries
 
 
+def load_audio_translations(
+    *, supplement_paths: tuple[Path, ...]
+) -> dict[str, str]:
+    """Collect the spoken-line translations, keyed by the normalized original.
+
+    ``load_audio_transcript`` can only backfill translations for file ids that
+    already appear in a transcript; this map covers the remaining lines the
+    game itself ships.
+    """
+    translations: dict[str, str] = {}
+    for supplement_path in supplement_paths:
+        for original, translated in _load_audio_translations_by_original(supplement_path).items():
+            translations.setdefault(audio_translation_lookup_key(original), translated)
+        translations.update(_load_audio_translations_by_normalized_original(supplement_path))
+    return translations
+
+
+def translated_text_for(original_text: str, translations: dict[str, str]) -> str | None:
+    """Look up a translation for a spoken line, aligning terminal punctuation."""
+    translated = translations.get(audio_translation_lookup_key(original_text))
+    if not translated:
+        return None
+    return _align_translation_punctuation(original_text, translated)
+
+
 def _load_audio_translations_by_original(path: Path) -> dict[str, str]:
     if not path.is_file():
         return {}
@@ -225,16 +260,16 @@ def _load_audio_translations_by_normalized_original(path: Path) -> dict[str, str
     if not isinstance(raw_translations, dict):
         return {}
     return {
-        _audio_translation_lookup_key(original): translated.strip()
+        audio_translation_lookup_key(original): translated.strip()
         for original, translated in raw_translations.items()
         if isinstance(original, str)
-        and _audio_translation_lookup_key(original)
+        and audio_translation_lookup_key(original)
         and isinstance(translated, str)
         and translated.strip()
     }
 
 
-def _audio_translation_lookup_key(text: str) -> str:
+def audio_translation_lookup_key(text: str) -> str:
     return " ".join(re.findall(r"[\w]+", text.casefold().replace("_", " ")))
 
 
@@ -474,8 +509,10 @@ def reference_status(path: Path) -> dict[str, object]:
     "BUNDLED_UNIT_VOICE_TRANSCRIPT_PATH",
     "BUNDLED_VOICE_TRANSLATION_PATH",
     "load_audio_transcript",
+    "load_audio_translations",
     "load_known_names",
     "reference_status",
     "sync_audio_transcript",
     "sync_known_names",
+    "translated_text_for",
 ]
diff --git a/src/ra2_explorer/semantic.py b/src/ra2_explorer/semantic.py
index 2846fe9..c7cec82 100644
--- a/src/ra2_explorer/semantic.py
+++ b/src/ra2_explorer/semantic.py
@@ -36,12 +36,13 @@
     localized_mixed_search_match,
     pinyin_search_aliases,
 )
+from ra2_explorer.reference_data import translated_text_for
 from ra2_explorer.storage import Database
 
 ENTITY_KINDS = ("vehicle", "infantry", "aircraft", "building")
 ENTITY_USAGES = ("buildable", "hero", "tech", "civilian", "scenario")
 UNAFFILIATED_SIDE = "unaffiliated"
-SEMANTIC_CATALOG_CACHE_IDENTITY = ("semantic-catalog-v23",)
+SEMANTIC_CATALOG_CACHE_IDENTITY = ("semantic-catalog-v24",)
 _PLANNING_SIDE_IDS = ("GDI", "Nod", "ThirdSide")
 _TYPE_SECTIONS = {
     "vehicle": "VehicleTypes",
@@ -750,13 +751,15 @@ def __init__(
         database: Database,
         reader: AssetReader,
         voice_transcripts: dict[str, dict[str, str]] | None = None,
+        voice_translations: dict[str, str] | None = None,
     ):
         self.database = database
         self.reader = reader
         self.voice_transcripts = voice_transcripts or {}
+        self.voice_translations = voice_translations or {}
         self._voice_transcript_revision = hashlib.sha256(
             json.dumps(
-                self.voice_transcripts,
+                (self.voice_transcripts, self.voice_translations),
                 ensure_ascii=False,
                 sort_keys=True,
                 separators=(",", ":"),
@@ -984,7 +987,7 @@ def list_media(
             "total": total,
             "kinds": [
                 {"kind": media_kind, "count": kind_counts.get(media_kind, 0)}
-                for media_kind in ("voice", "sound", "unknown")
+                for media_kind in ("voice", "sound", "music", "unknown")
             ],
             "groups": [
                 {"group": media_group, "count": count}
@@ -1639,13 +1642,14 @@ def _build(self, source_id: str) -> SemanticCatalog:
 
         warnings = []
         rules_assets = _edition_ini_inputs(
-            asset_index.by_name,
+            asset_index,
             "rules.ini",
             "rulesmd.ini",
         )
-        art_assets = _named_inputs(asset_index.by_name, ("art.ini", "artmd.ini"))
-        sound_assets = _named_inputs(asset_index.by_name, ("sound.ini", "soundmd.ini"))
-        eva_assets = _named_inputs(asset_index.by_name, ("eva.ini", "evamd.ini"))
+        art_assets = _named_inputs(asset_index, ("art.ini", "artmd.ini"))
+        sound_assets = _named_inputs(asset_index, ("sound.ini", "soundmd.ini"))
+        eva_assets = _named_inputs(asset_index, ("eva.ini", "evamd.ini"))
+        theme_assets = _named_inputs(asset_index, ("theme.ini", "thememd.ini"))
         csf_assets = sorted(
             (asset for asset in assets if asset["format"] == "csf"),
             key=_config_precedence,
@@ -1654,11 +1658,13 @@ def _build(self, source_id: str) -> SemanticCatalog:
         art = _merge_ini_inputs(self.reader, art_assets, warnings)
         sounds = _merge_ini_inputs(self.reader, sound_assets, warnings)
         eva = _merge_ini_inputs(self.reader, eva_assets, warnings)
+        themes = _merge_ini_inputs(self.reader, theme_assets, warnings)
         strings, voice_strings = _merge_csf_inputs(
             self.reader,
             csf_assets,
             warnings,
             self.voice_transcripts,
+            self.voice_translations,
         )
         audio_events = _build_audio_events(sounds, asset_index, voice_strings)
         country_definitions = _build_country_definitions(rules, strings)
@@ -1760,6 +1766,7 @@ def _build(self, source_id: str) -> SemanticCatalog:
             "art": tuple(_input_summary(asset) for asset in art_assets),
             "sound": tuple(_input_summary(asset) for asset in sound_assets),
             "eva": tuple(_input_summary(asset) for asset in eva_assets),
+            "theme": tuple(_input_summary(asset) for asset in theme_assets),
             "csf": tuple(_input_summary(asset) for asset in csf_assets),
         }
         if not rules_assets:
@@ -1772,6 +1779,8 @@ def _build(self, source_id: str) -> SemanticCatalog:
             audio_events,
             eva_events,
             voice_strings,
+            strings,
+            _theme_entries(themes),
         )
         return SemanticCatalog(
             source_id,
@@ -2359,12 +2368,44 @@ def _mission_event_slot(asset_stem: str, events: Iterable[str]) -> str:
     return "mission_dialogue"
 
 
+@dataclass(frozen=True, slots=True)
+class ThemeEntry:
+    stem: str
+    title: str | None
+
+
+def _theme_entries(themes: dict[str, dict[str, str]]) -> tuple[ThemeEntry, ...]:
+    """Bind THEME.INI soundtrack entries to their audio file stems.
+
+    Retail THEME.INI keeps a [Themes] index plus one section per track whose
+    Sound key names the audio file and whose Name key points at a CSF label.
+    """
+    entries: list[ThemeEntry] = []
+    seen: set[str] = set()
+    for section, values in themes.items():
+        if section.casefold() == "themes":
+            continue
+        sound = str(values.get("sound") or "").strip()
+        if not sound:
+            continue
+        stem = sound.rsplit(".", 1)[0]
+        folded = stem.casefold()
+        if not folded or folded in seen:
+            continue
+        seen.add(folded)
+        title = str(values.get("name") or "").strip() or None
+        entries.append(ThemeEntry(stem, title))
+    return tuple(entries)
+
+
 def _build_media_items(
     assets: list[dict[str, Any]],
     entities: tuple[GameEntity, ...],
     audio_events: dict[str, tuple[MediaSample, ...]],
     eva_events: tuple[MediaAssociation, ...],
     voice_strings: dict[str, VoiceText],
+    strings: dict[str, str] | None = None,
+    theme_entries: tuple[ThemeEntry, ...] = (),
 ) -> tuple[dict[str, object], ...]:
     entities_by_id = {entity.id.casefold(): entity for entity in entities}
     representatives: dict[str, dict[str, Any]] = {}
@@ -2381,6 +2422,7 @@ def _build_media_items(
             "asset": asset,
             "voice": False,
             "sound": False,
+            "music": False,
             "groups": set(),
             "texts": set(),
             "original_texts": set(),
@@ -2622,8 +2664,22 @@ def add_sample(
                 state["events"].add("PropagandaTruck")
                 state["slots"].add("ambient")
 
+    for entry in theme_entries:
+        state = states.get(f"{entry.stem}.wav".casefold()) or states.get(
+            f"{entry.stem}.aud".casefold()
+        )
+        if state is None:
+            continue
+        state["music"] = True
+        state["groups"].add("theme_music")
+        state["events"].add(entry.stem)
+        state["slots"].add("theme")
+        title = strings.get(str(entry.title).casefold()) if strings and entry.title else None
+        if title:
+            state["texts"].add(title.strip())
+
     for state in states.values():
-        if state["voice"] or state["sound"]:
+        if state["voice"] or state["sound"] or state["music"]:
             continue
         asset = state["asset"]
         stem = str(asset["display_name"]).rsplit(".", 1)[0].casefold()
@@ -2636,6 +2692,14 @@ def add_sample(
             state["events"].add(event)
             state["slots"].add("explosion")
             continue
+        # Tracks that THEME.INI omits or comments out still ship inside the
+        # soundtrack archives, so archive membership is the reliable signal.
+        if "theme.mix" in virtual_path or "thememd.mix" in virtual_path:
+            state["music"] = True
+            state["groups"].add("theme_music")
+            state["events"].add(stem.upper())
+            state["slots"].add("theme")
+            continue
         description = _LEGACY_INTERFACE_AUDIO_DESCRIPTIONS.get(stem)
         if (
             description is not None
@@ -2653,7 +2717,12 @@ def add_sample(
 
     items = []
     for state in states.values():
-        kind = "voice" if state["voice"] else "sound" if state["sound"] else "unknown"
+        kind = (
+            "voice" if state["voice"]
+            else "sound" if state["sound"]
+            else "music" if state["music"]
+            else "unknown"
+        )
         asset_stem = str(state["asset"]["display_name"]).rsplit(".", 1)[0].casefold()
         events = []
         seen_events: set[str] = set()
@@ -2681,6 +2750,7 @@ def add_sample(
             for group in refined_groups
             if (kind == "voice" and group.endswith("_voice"))
             or (kind == "sound" and group.endswith("_sound"))
+            or (kind == "music" and group.endswith("_music"))
         )
         if kind == "voice" and len(groups) > 1 and "other_voice" in groups:
             groups.remove("other_voice")
@@ -2750,22 +2820,39 @@ def add_sample(
 
 
 def _named_inputs(
-    by_name: dict[str, list[dict[str, Any]]],
+    asset_index: _AssetIndex,
     names: tuple[str, ...],
 ) -> list[dict[str, Any]]:
-    assets = []
+    assets: list[dict[str, Any]] = []
+    seen: set[str] = set()
     for name in names:
-        assets.extend(by_name.get(name.casefold(), ()))
+        candidates = list(asset_index.by_name.get(name.casefold(), ()))
+        if not candidates:
+            # Unnamed members are indexed under their CRC; resolve known
+            # configuration names through their MIX hashes as well.
+            try:
+                hashes = {ra2_mix_hash(name), classic_mix_hash(name)}
+            except ValueError:
+                continue
+            candidates = [
+                asset
+                for crc in hashes
+                for asset in asset_index.by_crc.get(crc, ())
+            ]
+        for asset in candidates:
+            if asset["id"] not in seen:
+                seen.add(asset["id"])
+                assets.append(asset)
     return sorted(assets, key=_config_precedence)
 
 
 def _edition_ini_inputs(
-    by_name: dict[str, list[dict[str, Any]]],
+    asset_index: _AssetIndex,
     base_name: str,
     expansion_name: str,
 ) -> list[dict[str, Any]]:
-    expansion = _named_inputs(by_name, (expansion_name,))
-    return expansion or _named_inputs(by_name, (base_name,))
+    expansion = _named_inputs(asset_index, (expansion_name,))
+    return expansion or _named_inputs(asset_index, (base_name,))
 
 
 def _merge_ini_inputs(
@@ -2793,6 +2880,7 @@ def _merge_csf_inputs(
     assets: list[dict[str, Any]],
     warnings: list[str],
     voice_transcripts: dict[str, dict[str, str]],
+    voice_translations: dict[str, str] | None = None,
 ) -> tuple[dict[str, str], dict[str, VoiceText]]:
     strings: dict[str, str] = {}
     voice_strings: dict[str, VoiceText] = {}
@@ -2809,6 +2897,10 @@ def _merge_csf_inputs(
                 strings[folded] = label.values[0].text
                 if folded.startswith("vox:"):
                     value = label.values[0]
+                    # Chinese community installs often leave the CSF language
+                    # header at 0; treat CJK text as the localized column so
+                    # packaged Chinese voice lines are not misfiled as English.
+                    is_localized = parsed.language == 9 or _contains_cjk(value.text)
                     for alias in _voice_aliases(folded[4:], value.extra):
                         current = voice_strings.get(alias)
                         original_text = current.original_text if current else None
@@ -2817,7 +2909,7 @@ def _merge_csf_inputs(
                             current.localized_text_origin if current else None
                         )
                         translated_text = current.translated_text if current else None
-                        if parsed.language == 9:
+                        if is_localized:
                             localized_text = value.text
                             localized_text_origin = "game"
                         elif parsed.language == 0:
@@ -2831,6 +2923,7 @@ def _merge_csf_inputs(
                             translated_text,
                         )
     _overlay_voice_transcripts(voice_strings, voice_transcripts)
+    _overlay_voice_translations(voice_strings, voice_translations or {})
     return strings, voice_strings
 
 
@@ -2873,6 +2966,41 @@ def _overlay_voice_transcripts(
         )
 
 
+def _overlay_voice_translations(
+    voice_strings: dict[str, VoiceText],
+    translations: dict[str, str],
+) -> None:
+    """Translate spoken lines the game ships but no transcript file lists.
+
+    ``load_audio_transcript`` can only backfill translations for file ids that
+    already have a transcript entry, so lines that only exist in the game CSF
+    would otherwise stay untranslated even when a translation is available.
+    """
+    if not translations:
+        return
+    for key, current in list(voice_strings.items()):
+        if current.translated_text or not current.original_text:
+            continue
+        translated_text = translated_text_for(current.original_text, translations)
+        if not translated_text:
+            continue
+        voice_strings[key] = VoiceText(
+            current.label,
+            translated_text if current.text == current.original_text else current.text,
+            current.original_text,
+            current.localized_text,
+            current.localized_text_origin,
+            translated_text,
+        )
+
+
+def _contains_cjk(text: str) -> bool:
+    return any(
+        0x2E80 <= ord(character) <= 0x9FFF or 0xF900 <= ord(character) <= 0xFAFF
+        for character in text
+    )
+
+
 def _voice_aliases(label_name: str, extra: str | None) -> tuple[str, ...]:
     aliases = [label_name.casefold()]
     for raw in re.split(r"[,;|\s]+", extra or ""):
diff --git a/src/ra2_explorer/video.py b/src/ra2_explorer/video.py
index 1270be9..0b7c3f8 100644
--- a/src/ra2_explorer/video.py
+++ b/src/ra2_explorer/video.py
@@ -6,11 +6,22 @@
 import uuid
 from pathlib import Path
 
+from PIL import Image
+
 from ra2_explorer.derived import DerivedStore
 from ra2_explorer.errors import Ra2ExplorerError
 from ra2_explorer.library import AssetReader
 from ra2_explorer.storage import Database
 
+# Posters skip the opening seconds because these movies fade in from black,
+# then walk deeper until something readable shows up. The last attempt has no
+# seek at all, so even a clip shorter than every seek still gets a poster.
+POSTER_SEEK_SECONDS: tuple[float, ...] = (1, 2, 4, 6, 8)
+# A frame where fewer than this share of pixels are brighter than the luma
+# floor counts as black and is not worth showing.
+POSTER_DARK_LUMA = 16
+POSTER_DARK_MAX_RATIO = 0.02
+
 
 class VideoTranscoder:
     """Create browser-compatible video derivatives without touching game files."""
@@ -45,59 +56,151 @@ def browser_video(self, asset_id: str) -> Path:
             _, source_path = self.reader.materialize(asset_id)
             output.parent.mkdir(parents=True, exist_ok=True)
             temporary = output.with_name(f".{output.stem}.{uuid.uuid4().hex}.tmp.mp4")
-            creation_flags = getattr(subprocess, "CREATE_NO_WINDOW", 0)
-            command = [
-                ffmpeg,
-                "-hide_banner",
-                "-loglevel",
-                "error",
-                "-nostdin",
-                "-y",
-                "-i",
-                str(source_path),
-                "-map",
-                "0:v:0",
-                "-map",
-                "0:a:0?",
-                "-c:v",
-                "libx264",
-                "-preset",
-                "veryfast",
-                "-crf",
-                "20",
-                "-pix_fmt",
-                "yuv420p",
-                "-c:a",
-                "aac",
-                "-b:a",
-                "160k",
-                "-movflags",
-                "+faststart",
-                str(temporary),
-            ]
-            try:
-                result = subprocess.run(
-                    command,
-                    check=False,
-                    stdin=subprocess.DEVNULL,
-                    stdout=subprocess.DEVNULL,
-                    stderr=subprocess.PIPE,
-                    timeout=300,
-                    creationflags=creation_flags,
-                )
-                if result.returncode != 0 or not temporary.is_file():
-                    detail = result.stderr.decode("utf-8", errors="replace").strip()[-600:]
-                    raise Ra2ExplorerError(f"视频转换失败：{detail or 'FFmpeg 未生成输出'}")
+            self._run_ffmpeg(
+                [
+                    ffmpeg,
+                    "-hide_banner",
+                    "-loglevel",
+                    "error",
+                    "-nostdin",
+                    "-y",
+                    "-i",
+                    str(source_path),
+                    "-map",
+                    "0:v:0",
+                    "-map",
+                    "0:a:0?",
+                    "-c:v",
+                    "libx264",
+                    "-preset",
+                    "veryfast",
+                    "-crf",
+                    "20",
+                    "-pix_fmt",
+                    "yuv420p",
+                    "-c:a",
+                    "aac",
+                    "-b:a",
+                    "160k",
+                    "-movflags",
+                    "+faststart",
+                    str(temporary),
+                ],
+                output,
+                "视频转换失败",
+            )
+        return output
+
+    def poster_frame(self, asset_id: str) -> Path:
+        """Extract one frame so video assets can show a preview before transcoding."""
+        asset = self.database.get_asset(asset_id)
+        if asset["format"] != "video":
+            raise Ra2ExplorerError("该资产不是可转换的视频")
+        source = self.database.get_source(str(asset["source_id"]))
+        output = self.derived.artifact_path(
+            "video",
+            source_id=source["id"],
+            revision=source.get("scanned_at") or source["created_at"],
+            identity=(asset["id"], "poster-v2"),
+            extension="png",
+        )
+        if output.is_file():
+            return output
+        with self._lock_for(output):
+            if output.is_file():
+                return output
+            ffmpeg = shutil.which("ffmpeg")
+            if ffmpeg is None:
+                raise Ra2ExplorerError("未找到 FFmpeg，无法生成视频封面")
+            _, source_path = self.reader.materialize(asset_id)
+            output.parent.mkdir(parents=True, exist_ok=True)
+            temporary = output.with_name(f".{output.stem}.{uuid.uuid4().hex}.tmp.png")
+            self._capture_poster(ffmpeg, source_path, output, temporary)
+        return output
+
+    def _capture_poster(
+        self,
+        ffmpeg: str,
+        source_path: Path,
+        output: Path,
+        temporary: Path,
+    ) -> None:
+        """Seek progressively deeper until the poster is not a black frame."""
+        attempts: list[float | None] = [*POSTER_SEEK_SECONDS, None]
+        detail = "FFmpeg 未生成输出"
+        for seconds in attempts:
+            command = [ffmpeg, "-hide_banner", "-loglevel", "error", "-nostdin", "-y"]
+            if seconds is not None:
+                command += ["-ss", f"{seconds:g}"]
+            command += ["-i", str(source_path), "-frames:v", "1", str(temporary)]
+            ok, attempt_detail = self._extract_frame(command)
+            if ok and (seconds is None or not _frame_is_dark(temporary)):
                 self.derived.commit_file(output, temporary)
-            except subprocess.TimeoutExpired as error:
-                raise Ra2ExplorerError("视频转换超过 5 分钟，已停止") from error
-            finally:
                 temporary.unlink(missing_ok=True)
-        return output
+                return
+            temporary.unlink(missing_ok=True)
+            detail = attempt_detail or detail
+        raise Ra2ExplorerError(f"视频封面生成失败：{detail or '未找到可用的画面帧'}")
+
+    def _extract_frame(self, command: list[str]) -> tuple[bool, str]:
+        """Extract a single frame without publishing it."""
+        try:
+            result = subprocess.run(
+                command,
+                check=False,
+                stdin=subprocess.DEVNULL,
+                stdout=subprocess.DEVNULL,
+                stderr=subprocess.PIPE,
+                timeout=120,
+                creationflags=getattr(subprocess, "CREATE_NO_WINDOW", 0),
+            )
+        except subprocess.TimeoutExpired:
+            return False, "单帧抽取超过 120 秒已停止"
+        detail = result.stderr.decode("utf-8", errors="replace").strip()[-600:]
+        if result.returncode != 0:
+            return False, detail or f"FFmpeg 退出码 {result.returncode}"
+        if not Path(command[-1]).is_file():
+            # Seeking past the end of a very short clip yields no frame.
+            return False, detail or "FFmpeg 未生成输出"
+        return True, ""
+
+    def _run_ffmpeg(self, command: list[str], output: Path, failure: str) -> None:
+        temporary = Path(command[-1])
+        try:
+            result = subprocess.run(
+                command,
+                check=False,
+                stdin=subprocess.DEVNULL,
+                stdout=subprocess.DEVNULL,
+                stderr=subprocess.PIPE,
+                timeout=300,
+                creationflags=getattr(subprocess, "CREATE_NO_WINDOW", 0),
+            )
+            if result.returncode != 0 or not temporary.is_file():
+                detail = result.stderr.decode("utf-8", errors="replace").strip()[-600:]
+                raise Ra2ExplorerError(f"{failure}：{detail or 'FFmpeg 未生成输出'}")
+            self.derived.commit_file(output, temporary)
+        except subprocess.TimeoutExpired as error:
+            raise Ra2ExplorerError(f"{failure}：超过 5 分钟已停止") from error
+        finally:
+            temporary.unlink(missing_ok=True)
 
     def _lock_for(self, path: Path) -> threading.Lock:
         with self._locks_guard:
             return self._locks.setdefault(path, threading.Lock())
 
 
+def _frame_is_dark(path: Path) -> bool:
+    """Return True when a frame is essentially black, such as a fade-in."""
+    try:
+        with Image.open(path) as image:
+            histogram = image.convert("L").histogram()
+    except OSError:
+        return True
+    total = sum(histogram)
+    if not total:
+        return True
+    return sum(histogram[POSTER_DARK_LUMA:]) / total < POSTER_DARK_MAX_RATIO
+
+
 __all__ = ["VideoTranscoder"]
diff --git a/tests/test_api.py b/tests/test_api.py
index 2fb678a..2aa4693 100644
--- a/tests/test_api.py
+++ b/tests/test_api.py
@@ -3,6 +3,7 @@
 import io
 import json
 from pathlib import Path
+from types import SimpleNamespace
 
 from fastapi.testclient import TestClient
 from PIL import Image
@@ -13,9 +14,12 @@
     _composite_focus_bounds,
     _crop_transparent_preview,
     _default_entity_operation_samples,
+    _select_palette,
     create_app,
 )
+from ra2_explorer.codecs.pal import PLACEHOLDER_MAGENTA
 from ra2_explorer.config import Settings
+from ra2_explorer.reference_data import audio_translation_lookup_key
 from ra2_explorer.semantic import (
     GameEntity,
     MediaAssociation,
@@ -35,6 +39,7 @@
     _entity_usage,
     _index_assets,
     _overlay_voice_transcripts,
+    _overlay_voice_translations,
     _resolve_components,
     _shp_unit_body_playbacks,
     _standalone_sound_group,
@@ -863,6 +868,96 @@ def test_verified_unit_intro_overrides_csf_unit_name_text() -> None:
     assert voice_strings["csofu39"].localized_text_origin == "game"
 
 
+def test_translation_backfill_covers_lines_without_transcript_entries() -> None:
+    voice_strings = {
+        "cevau06": VoiceText(
+            "VOX:CEVAU06",
+            "The V3 is a powerful long-range artillery weapon.",
+            "The V3 is a powerful long-range artillery weapon.",
+            None,
+            "game",
+            None,
+        )
+    }
+
+    _overlay_voice_translations(
+        voice_strings,
+        {
+            audio_translation_lookup_key(
+                "The V3 is a powerful long-range artillery weapon."
+            ): "V3 是威力强大的远程火炮。"
+        },
+    )
+
+    entry = voice_strings["cevau06"]
+    assert entry.translated_text == "V3 是威力强大的远程火炮。"
+    assert entry.original_text == "The V3 is a powerful long-range artillery weapon."
+    assert entry.text == "V3 是威力强大的远程火炮。"
+    assert entry.localized_text_origin == "game"
+
+
+def test_translation_backfill_keeps_existing_and_nonverbal_entries() -> None:
+    voice_strings = {
+        "covered": VoiceText("VOX:COVERED", "Line", "Line", None, "game", "已有译文"),
+        "nonverbal": VoiceText("VOX:NONVERBAL", "<阵亡声>", None, None, None, "<阵亡声>"),
+    }
+
+    _overlay_voice_translations(
+        voice_strings,
+        {
+            audio_translation_lookup_key("Line"): "不应覆盖",
+            audio_translation_lookup_key("<阵亡声>"): "不应出现",
+        },
+    )
+
+    assert voice_strings["covered"].translated_text == "已有译文"
+    assert voice_strings["nonverbal"].translated_text == "<阵亡声>"
+
+
+def test_palette_selection_avoids_magenta_placeholders() -> None:
+    def palette_data_with(colors: dict[int, tuple[int, int, int]]) -> bytes:
+        raw = bytearray()
+        for index in range(256):
+            red, green, blue = colors.get(index, (index, index, index))
+            raw += bytes((min(63, red >> 2), min(63, green >> 2), min(63, blue >> 2)))
+        return bytes(raw)
+
+    # 单位调色板把 204-239 填成洋红占位，地形调色板则是真实颜色。
+    unit = palette_data_with({index: PLACEHOLDER_MAGENTA for index in range(204, 240)})
+    iso = palette_data_with({204: (68, 56, 52)})
+    palettes = [
+        {"id": "unit-ubn", "display_name": "UNITUBN.PAL"},
+        {"id": "unit-tem", "display_name": "unittem.pal"},
+        {"id": "iso-ubn", "display_name": "ISOUBN.PAL"},
+        {"id": "iso-tem", "display_name": "isotem.pal"},
+    ]
+    data = {
+        "unit-ubn": unit,
+        "unit-tem": unit,
+        "iso-ubn": iso,
+        "iso-tem": palette_data_with({}),
+    }
+    services = SimpleNamespace(
+        database=SimpleNamespace(palette_assets=lambda source_id: palettes),
+        reader=SimpleNamespace(read=lambda asset_id: (None, data[asset_id])),
+    )
+    asset = {
+        "id": "urban-object",
+        "source_id": "source",
+        "format": "shp",
+        "extension": "pak",
+        "virtual_path": "ra2md.mix/URBANN.MIX::00102:0ABB14BC-G.PAK",
+    }
+
+    # 缺省时沿用单位调色板；画面用到占位索引时改用地形调色板。
+    assert _select_palette(services, asset, None, None).colors[204] == PLACEHOLDER_MAGENTA
+    scored = _select_palette(services, asset, None, None, used_indices={204, 130})
+    assert scored.colors[204] == (68, 56, 52)
+    # 明确的调色板用途（如 art.ini 声明的单位动画）仍然优先单位调色板。
+    forced = _select_palette(services, asset, None, "unit", used_indices={204})
+    assert forced.colors[204] == PLACEHOLDER_MAGENTA
+
+
 def test_eva_media_groups_separate_missions_and_nonverbal_prompts() -> None:
     def asset(name: str) -> dict[str, object]:
         return {
@@ -1675,13 +1770,16 @@ def test_expansion_rules_replace_the_base_edition() -> None:
         "storage_kind": "mix",
     }
 
+    def index_for(by_name: dict[str, list[dict[str, object]]]) -> _AssetIndex:
+        return _AssetIndex(by_name=dict(by_name), by_crc={})
+
     assert _edition_ini_inputs(
-        {"rules.ini": [base], "rulesmd.ini": [expansion]},
+        index_for({"rules.ini": [base], "rulesmd.ini": [expansion]}),
         "rules.ini",
         "rulesmd.ini",
     ) == [expansion]
     assert _edition_ini_inputs(
-        {"rules.ini": [base]},
+        index_for({"rules.ini": [base]}),
         "rules.ini",
         "rulesmd.ini",
     ) == [base]
diff --git a/tests/test_library_scan.py b/tests/test_library_scan.py
new file mode 100644
index 0000000..b42bb52
--- /dev/null
+++ b/tests/test_library_scan.py
@@ -0,0 +1,118 @@
+from __future__ import annotations
+
+from pathlib import Path
+
+from ra2_explorer.codecs.mix import build_mix
+from ra2_explorer.codecs.sniff import sniff_format
+from ra2_explorer.library import SourceLibrary
+from ra2_explorer.storage import Database
+
+
+def test_unnamed_encrypted_nested_mix_is_traversed(tmp_path: Path) -> None:
+    source_dir = tmp_path / "game"
+    source_dir.mkdir()
+    inner = build_mix(
+        (
+            ("fixtone.wav", b"\x00" * 64),
+            ("fixture.ini", b"[General]\nName=Fixture\n"),
+        ),
+        encrypted=True,
+    )
+    decoy = b"A" * 2048
+    outer = build_mix((("sideblob.dat", inner), ("decoy.bin", decoy)))
+    (source_dir / "outer.mix").write_bytes(outer)
+
+    database = Database(tmp_path / "index.db")
+    source = SourceLibrary(
+        database, ("fixtone.wav", "fixture.ini")
+    ).import_source(source_dir)
+    items = database.list_assets(source_id=str(source["id"]), limit=500)["items"]
+
+    formats = {asset["display_name"]: asset["format"] for asset in items}
+    assert formats["fixtone.wav"] == "wav"
+    assert formats["fixture.ini"] == "ini"
+    # The decoy has no resolvable name and must not be mistaken for a MIX.
+    assert sum(asset_format == "binary" for asset_format in formats.values()) == 1
+    nested_mix_names = [
+        name for name, asset_format in formats.items()
+        if asset_format == "mix" and name.startswith("crc_")
+    ]
+    assert len(nested_mix_names) == 1
+
+
+def test_sniff_detects_bink_video() -> None:
+    assert sniff_format(b"BIKi" + b"\x00" * 64, None) == "video"
+
+
+def _csf_bytes(language: int, labels: tuple[tuple[str, str, str | None], ...]) -> bytes:
+    import struct
+
+    output = bytearray(b" FSC")
+    output.extend(struct.pack("<IIIII", 3, len(labels), len(labels), 0, language))
+    for name, text, extra in labels:
+        encoded_name = name.encode("ascii")
+        output.extend(b" LBL")
+        output.extend(struct.pack("<II", 1, len(encoded_name)))
+        output.extend(encoded_name)
+        output.extend(b"WRTS" if extra else b" RTS")
+        units = text.encode("utf-16-le")
+        output.extend(struct.pack("<I", len(units) // 2))
+        output.extend(byte ^ 0xFF for byte in units)
+        if extra:
+            encoded_extra = extra.encode("ascii")
+            output.extend(struct.pack("<I", len(encoded_extra)))
+            output.extend(encoded_extra)
+    return bytes(output)
+
+
+class _FakeReader:
+    def __init__(self, payloads: dict[str, bytes]) -> None:
+        self._payloads = payloads
+
+    def read(self, asset_id: str) -> tuple[str, bytes]:
+        return asset_id, self._payloads[asset_id]
+
+
+def test_csf_voice_text_falls_back_to_cjk_detection() -> None:
+    from ra2_explorer.semantic import _merge_csf_inputs
+
+    english = _csf_bytes(
+        0,
+        (("VOX:apocsea", "Soviet power supreme", "vaposea"),),
+    )
+    chinese = _csf_bytes(
+        0,
+        (("VOX:apocsea", "苏维埃联盟万岁", "vaposea"),),
+    )
+    assets = [
+        {"id": "csf-english", "display_name": "ra2.csf", "virtual_path": "ra2.csf"},
+        {"id": "csf-chinese", "display_name": "ra2zh.csf", "virtual_path": "ra2zh.csf"},
+    ]
+    reader = _FakeReader({"csf-english": english, "csf-chinese": chinese})
+
+    _, voice_strings = _merge_csf_inputs(reader, assets, [], {})
+
+    voice = voice_strings["vaposea"]
+    assert voice.original_text == "Soviet power supreme"
+    assert voice.localized_text == "苏维埃联盟万岁"
+
+
+def test_unnamed_audio_bag_pair_is_expanded(tmp_path: Path) -> None:
+    import struct
+
+    source_dir = tmp_path / "game"
+    source_dir.mkdir()
+    bag_payload = b"\x00" * 64
+    idx_payload = b"GABA" + struct.pack(
+        "<II16sIIIII", 2, 1, b"vaposea", 0, len(bag_payload), 22050, 0x02, 0
+    )
+    outer = build_mix((("audio.idx", idx_payload), ("audio.bag", bag_payload)))
+    (source_dir / "audio.mix").write_bytes(outer)
+
+    database = Database(tmp_path / "index.db")
+    source = SourceLibrary(database, ()).import_source(source_dir)
+    items = database.list_assets(source_id=str(source["id"]), limit=500)["items"]
+
+    bag_audio = [asset for asset in items if asset["format"] == "bag_audio"]
+    assert len(bag_audio) == 1
+    assert bag_audio[0]["display_name"] == "vaposea.wav"
diff --git a/tests/test_video.py b/tests/test_video.py
index 8b72d3d..af73345 100644
--- a/tests/test_video.py
+++ b/tests/test_video.py
@@ -3,17 +3,18 @@
 from pathlib import Path
 from types import SimpleNamespace
 
+from PIL import Image
+
 from ra2_explorer.derived import DerivedStore
 from ra2_explorer.library import AssetReader, SourceLibrary
 from ra2_explorer.storage import Database
-from ra2_explorer.video import VideoTranscoder
+from ra2_explorer.video import POSTER_SEEK_SECONDS, VideoTranscoder
 
 
-def test_video_transcode_is_on_demand_cached_and_derived(tmp_path: Path, monkeypatch) -> None:
+def _transcoder_with_video(tmp_path: Path):
     source_dir = tmp_path / "game"
     source_dir.mkdir()
-    original = b"BIK synthetic test payload"
-    (source_dir / "intro.bik").write_bytes(original)
+    (source_dir / "intro.bik").write_bytes(b"BIK synthetic test payload")
     database = Database(tmp_path / "index.db")
     source = SourceLibrary(database, ()).import_source(source_dir)
     asset = database.list_assets(
@@ -23,6 +24,17 @@ def test_video_transcode_is_on_demand_cached_and_derived(tmp_path: Path, monkeyp
     derived = DerivedStore(tmp_path / "RA2MD-Ext")
     reader = AssetReader(database, derived)
     transcoder = VideoTranscoder(database, reader, derived)
+    return transcoder, str(asset["id"]), derived
+
+
+def _write_frame(path: Path, luma: int) -> None:
+    Image.new("L", (4, 4), luma).save(path)
+
+
+def test_video_transcode_is_on_demand_cached_and_derived(tmp_path: Path, monkeypatch) -> None:
+    transcoder, asset_id, derived = _transcoder_with_video(tmp_path)
+    source_dir = tmp_path / "game"
+    original = (source_dir / "intro.bik").read_bytes()
     calls = []
 
     monkeypatch.setattr("ra2_explorer.video.shutil.which", lambda _name: "ffmpeg.exe")
@@ -34,8 +46,8 @@ def run(command, **_kwargs):
 
     monkeypatch.setattr("ra2_explorer.video.subprocess.run", run)
 
-    first = transcoder.browser_video(str(asset["id"]))
-    second = transcoder.browser_video(str(asset["id"]))
+    first = transcoder.browser_video(asset_id)
+    second = transcoder.browser_video(asset_id)
 
     assert first == second
     assert first.read_bytes() == b"browser mp4"
@@ -43,3 +55,95 @@ def run(command, **_kwargs):
     assert (source_dir / "intro.bik").read_bytes() == original
     assert len(calls) == 1
     assert "-nostdin" in calls[0]
+
+
+def test_video_poster_is_extracted_once_and_cached(tmp_path: Path, monkeypatch) -> None:
+    transcoder, asset_id, derived = _transcoder_with_video(tmp_path)
+    calls = []
+
+    monkeypatch.setattr("ra2_explorer.video.shutil.which", lambda _name: "ffmpeg.exe")
+
+    def run(command, **_kwargs):
+        calls.append(command)
+        _write_frame(Path(command[-1]), 200)
+        return SimpleNamespace(returncode=0, stderr=b"")
+
+    monkeypatch.setattr("ra2_explorer.video.subprocess.run", run)
+
+    first = transcoder.poster_frame(asset_id)
+    second = transcoder.poster_frame(asset_id)
+
+    assert first == second
+    assert first.suffix == ".png"
+    assert first.is_relative_to(derived.root)
+    assert len(calls) == 1
+    # Skips the opening fade-in and stops after a single frame.
+    assert calls[0][calls[0].index("-ss") + 1] == "1"
+    assert calls[0][calls[0].index("-frames:v") + 1] == "1"
+
+
+def test_video_poster_walks_deeper_past_black_frames(tmp_path: Path, monkeypatch) -> None:
+    transcoder, asset_id, _ = _transcoder_with_video(tmp_path)
+    calls = []
+
+    monkeypatch.setattr("ra2_explorer.video.shutil.which", lambda _name: "ffmpeg.exe")
+
+    def run(command, **_kwargs):
+        calls.append(command)
+        _write_frame(Path(command[-1]), 0)
+        return SimpleNamespace(returncode=0, stderr=b"")
+
+    monkeypatch.setattr("ra2_explorer.video.subprocess.run", run)
+
+    poster = transcoder.poster_frame(asset_id)
+
+    # Every seek lands on a black frame, so it falls through to the unseeked
+    # first frame instead of giving up.
+    assert len(calls) == len(POSTER_SEEK_SECONDS) + 1
+    assert all("-ss" in call for call in calls[:-1])
+    assert "-ss" not in calls[-1]
+    assert poster.is_file()
+
+
+def test_video_poster_survives_clips_shorter_than_the_seeks(
+    tmp_path: Path, monkeypatch
+) -> None:
+    transcoder, asset_id, _ = _transcoder_with_video(tmp_path)
+    calls = []
+
+    monkeypatch.setattr("ra2_explorer.video.shutil.which", lambda _name: "ffmpeg.exe")
+
+    def run(command, **_kwargs):
+        calls.append(command)
+        if "-ss" in command:
+            # Seeking past the end of a short clip produces no output.
+            return SimpleNamespace(returncode=0, stderr=b"")
+        _write_frame(Path(command[-1]), 200)
+        return SimpleNamespace(returncode=0, stderr=b"")
+
+    monkeypatch.setattr("ra2_explorer.video.subprocess.run", run)
+
+    poster = transcoder.poster_frame(asset_id)
+
+    assert all("-ss" in call for call in calls[:-1])
+    assert "-ss" not in calls[-1]
+    assert poster.is_file()
+
+
+def test_video_poster_reports_failure_when_nothing_is_extracted(
+    tmp_path: Path, monkeypatch
+) -> None:
+    import pytest
+
+    transcoder, asset_id, _ = _transcoder_with_video(tmp_path)
+    from ra2_explorer.errors import Ra2ExplorerError
+
+    monkeypatch.setattr("ra2_explorer.video.shutil.which", lambda _name: "ffmpeg.exe")
+
+    def run(command, **_kwargs):
+        return SimpleNamespace(returncode=1, stderr=b"boom")
+
+    monkeypatch.setattr("ra2_explorer.video.subprocess.run", run)
+
+    with pytest.raises(Ra2ExplorerError, match="视频封面生成失败"):
+        transcoder.poster_frame(asset_id)