ai-media-editor — local AI media editor (Video · Audio · Podcast)

English · Deutsch

ai-media-editor — local AI media editor (Video · Audio · Podcast)

Note

AI / Agent Native Integration: ai-media-editor is specifically engineered for autonomous coding agents (Claude Code, Gemini/Antigravity, Codex). It provides deterministic project preparation, Scribe JSON schema generation, and timestamped frame contact-sheets so LLMs can visually inspect and cut media locally without third-party SaaS dependencies.

Quick Navigation

#	Section	Description
01	Key Capabilities	Core orchestrator capabilities, local STT routing, and agent tooling stack
02	Visual Architecture Flowchart	5-tier system topology diagram from raw media intake to deterministic export
03	End-to-End Execution Lifecycle Sequence	Sequence diagram detailing preflight, STT, frame contact sheets, and cut execution
04	Target Personas & Discoverability	4 core user profiles (`[PERSONA-01]`–`[PERSONA-04]`) and high-intent SEO queries
05	Comparative Matrix vs. Alternatives	10-dimension feature matrix vs Cloud SaaS, Heavy NLEs, Raw CLI, and Hosted APIs
06	The 8 Production Usecases	Comprehensive matrix of audio, video, explainer, and advertising workflows
07	Getting Started & Setup	Prerequisites, `<TOOLS_ROOT>` configuration, and environment initialization
08	CLI Reference & Commands	Step-by-step CLI commands for `doctor`, `modes`, `prepare`, and `frames`
09	Motion Graphics & Music Synthesis	Hyperframes styling and offline NumPy waveform storyline scoring
10	Governance & Runtime Invariants	10 non-negotiable architectural guarantees (`INV-LOCAL-01` to `INV-SLA-10`)
11	Security & Privacy SLA	100% local-first zero-egress posture, 48h initial response, and 5-day triage SLA
12	Sibling Projects & Ecosystem Matrix	16 federated tools across `ellmos-ai`, `dev-bricks`, `file-bricks`, and `open-bricks`
13	Quality Gates & Testing	Automated test suite, Ruff linting, compileall, and console window suppression
14	Machine-Readable Context (`llms.txt`)	LLM crawler manifest, search indexing keywords, and agent ingestion rules
15	Repository Structure	Detailed folder taxonomy across code, tools, stt, tests, and projects
16	Changelog & Releases	Release history, version milestones, and Pfad B discoverability updates
17	Level 1 SBOM & Third-Party Licenses	Permissive open-source inventory, RunAsInvoker guarantee, and Level 1 SBOM
18	Statutory Notice & Liability Limitation	MIT license, § 521 BGB Gefälligkeitsrecht disclaimer, and liability limits
19	Sister Project: `clip-storyboard-director`	The Director half of the Director/Cutter duo, bundled as a pinned optional dependency

1. Key Capabilities

Use an AI coding agent (e.g. Claude Code, Gemini, Codex) as a video/podcast editor — with local transcription instead of ElevenLabs Scribe. The orchestrator (editor.py) handles deterministic preparation (routing to the optimal STT engine/compute, producing schema-valid Scribe JSON, and packing takes); creative cutting, animation, and composition work is driven by the agent.

A three-tool stack:

video-use — cuts based on word-level transcripts (removes pauses/stumbles).
Hyperframes — HTML/CSS/JS → MP4 animations and motion graphics.
frontend-design skill — generates motion graphics / branding assets.

…with ElevenLabs Scribe transcription replaced by local engines. The default is faster-whisper plus text-based LLM speaker assignment for conversations; WhisperX is an optional acoustic-diarization engine. Compute is local by default, with optional remote-host-primary, local-fallback routing. The replacement writes the exact Scribe fields used by video-use, so downstream helpers run completely unpatched.

2. Visual Architecture Flowchart

flowchart TB
    subgraph Intake ["Layer 1: Input Channels & Media Intake"]
        V["Raw Video Recording (.mp4/.mov)"]
        A["Raw Audio Recording (.wav/.m4a)"]
        S["Storyline JSON (Narrative & Emotion)"]
    end

    subgraph Preflight ["Layer 2: Orchestrator & Preflight Inspection"]
        DOC["editor.py doctor<br/>Environment & Tooling Verification"]
        MODES["editor.py modes<br/>8 Specialized Media Use Cases"]
        PREP["editor.py prepare<br/>Project Staging & Metadata Contract"]
    end

    subgraph Engines ["Layer 3: Processing & Synthesis Engines"]
        FW["faster-whisper / WhisperX<br/>100% Local STT & Word Alignment"]
        DIAR["ai_media_editor/stt/diarize_llm.py<br/>Text-based Multi-Speaker Diarization"]
        MUSIC["ai_media_editor/tools/compose_music.py<br/>Procedural NumPy Waveform Synthesis"]
        HF["Hyperframes & hide-windows.cjs<br/>HTML/CSS/JS -> MP4 Motion Graphics"]
    end

    subgraph Agentic ["Layer 4: Agentic Interaction Layer"]
        SCRIBE["Scribe JSON Schema<br/>Word-level Timing for video-use"]
        FRAMES["ai_media_editor/tools/frame_view.py<br/>Timestamped Contact-Sheets"]
        CUTS["ai_media_editor/tools/cut_view.py<br/>Pause & Inflection Cut Candidates"]
        LLM["AI Coding Agent<br/>(Claude Code / Gemini / Codex)"]
    end

    subgraph Export ["Layer 5: Deterministic Export & Artefacts"]
        OUT_V["Rendered Master Video (.mp4)"]
        OUT_A["Clean Speaker-Separated Audio (.wav)"]
        OUT_M["Standard MIDI File (.mid) & Score"]
    end

    Intake --> Preflight
    Preflight --> Engines
    Engines --> Agentic
    Agentic --> Export

3. End-to-End Execution Lifecycle Sequence

sequenceDiagram
    autonumber
    actor Dev as User / AI Agent
    participant Orch as Orchestrator (editor.py)
    participant STT as Local STT Engine
    participant FS as Local Filesystem (projects/)
    participant Visual as Frame View & Cut View
    participant Render as Hyperframes & ffmpeg

    Dev->>Orch: editor.py doctor
    Orch-->>Dev: Preflight verified (ffmpeg, venv, Node)
    Dev->>Orch: editor.py prepare "<media>" --mode <1-8>
    Orch->>STT: Stream audio track (faster-whisper / WhisperX)
    STT-->>Orch: Word-level timestamps & speaker segments
    Orch->>FS: Write schema-valid Scribe JSON & packed takes
    Dev->>Orch: editor.py frames <project> --contact-sheet
    Orch->>Visual: Sample video at regular keyframes
    Visual->>FS: Generate timestamped contact-sheet PNGs
    Dev->>Visual: cut_view.py (find speech pauses & stumbles)
    Visual-->>Dev: Return pause cut candidate intervals
    Dev->>Render: Formulate cut list & motion graphics
    Render->>FS: Render final clean MP4 / WAV with zero network egress
    FS-->>Dev: Deliver production media artifact

4. Target Personas & Discoverability

ai-media-editor solves key workflow bottlenecks for modern AI media automation across four distinct profiles:

[PERSONA-01] Autonomous AI Coding Agent Developers & Swarm Architects

Profile: Software engineers designing multi-agent workflows (Claude Code, Gemini/Antigravity, Codex CLI, Kimi Code) for autonomous media production.
Pain Point: Proprietary cloud STT services impose API rate limits, require brittle browser automation, and leak confidential unreleased media to external vendors.
Solution: Pure CLI orchestration with deterministic Scribe JSON schema generation, pause cut candidates, and timestamped contact sheets directly ingestible by multimodal LLMs.
High-Intent Query: "agent driven video editor with local transcription"

[PERSONA-02] Local-First Podcasters & Content Creators

Profile: Independent podcasters, interviewers, and creators managing weekly audio/video releases.
Pain Point: Exorbitant monthly subscription costs for cloud editing SaaS (Descript, Riverside) and slow upload times for multi-gigabyte recordings.
Solution: 100% offline faster-whisper transcription, speaker-separated audio tracks, automated pause trimming, and procedural animated cover video generation.
High-Intent Query: "local first podcast editing AI agent workflow"

[PERSONA-03] AI Video & Motion Graphics Engineers

Profile: Front-end developers and motion designers automating branded visual video assets.
Pain Point: Traditional NLEs (Premiere, After Effects) lack clean programmatic interfaces and cannot render animations directly from CSS/HTML design tokens.
Solution: Direct Hyperframes integration converting HTML/CSS/JavaScript components into 60fps MP4 motion graphics with suppressed Windows console flashing.
High-Intent Query: "Hyperframes motion graphics podcast editor"

[PERSONA-04] Enterprise Media Security & Compliance Officers

Profile: Compliance and information security officers in regulated industries (healthcare, legal, banking).
Pain Point: Legal prohibitions (GDPR Art. 28, HIPAA, confidential M&A data) against uploading proprietary company recordings to cloud-hosted transcription providers.
Solution: Verifiable zero-network-egress guarantee (INV-LOCAL-01), unprivileged RunAsInvoker non-elevation (INV-RUNAS-02), Level 1 SBOM, and a formal 48-hour vulnerability response SLA.
High-Intent Query: "zero egress video transcription and cutting"

5. Comparative Matrix vs. Alternatives

The following 10-dimension matrix compares ai-media-editor against legacy and competing approaches mapped to our non-negotiable runtime invariants:

Dimension & Invariant	ai-media-editor	Cloud SaaS (Descript / ElevenLabs)	Commercial NLEs (Premiere / DaVinci)	Raw CLI / Shell (FFmpeg)	Hosted APIs (Whisper API / Google STT)
1. Local-First & Zero Egress (`INV-LOCAL-01`)	✅ 100% Offline Local	❌ Mandatory Cloud Upload	⚠️ Local (Proprietary Telemetry)	✅ 100% Offline Local	❌ External Cloud Processing
2. Unprivileged User-Mode (`INV-RUNAS-02`)	✅ `RunAsInvoker` Standard	⚠️ Browser / Helper App	⚠️ Admin Installer / Daemon	✅ User Privileges	✅ API Client Level
3. Non-Destructive Source Safety (`INV-PREV-03`)	✅ Read-Only Media Input	⚠️ Cloud Media Duplication	✅ Non-Destructive Projects	⚠️ Easy Overwrite Accident	✅ Input Media Preserved
4. Deterministic Synthesis (`INV-DETERM-04`)	✅ Scribe JSON & Seeded Score	❌ Stochastic Cloud Drift	⚠️ Complex Project State	⚠️ Unstructured Outputs	❌ Non-Deterministic Latency
5. Anti-Traversal Containment (`INV-BOUNDARY-05`)	✅ Strict Project Jail	⚠️ Hosted Cloud Bucket	⚠️ Arbitrary Filesystem	❌ No Path Validation	✅ Cloud Tenant Boundary
6. Window Flashing Suppression (`INV-SUBPROC-06`)	✅ `windowsHide` Preloaded	➖ N/A (Web UI)	➖ N/A (Desktop GUI)	❌ Disruptive Console Bursts	➖ N/A (HTTP Requests)
7. Cross-Platform Parity (`INV-PARITY-07`)	✅ Windows, Linux, macOS	⚠️ Browser-Dependent	❌ OS-Specific Features	✅ POSIX / Windows	✅ Cross-Platform HTTP
8. Multi-Agent Concurrency (`INV-SYNC-08`)	✅ Lock Discipline & Clean Git	❌ Single-User SaaS Session	❌ Single-Editor Project Locks	⚠️ Concurrency Conflicts	⚠️ Rate Limiting Bottlenecks
9. Agent & LLM Readability (`INV-DOCS-09`)	✅ Scribe JSON, Contact Sheets, `llms.txt`	❌ Closed Binary / HTML DOM	❌ Complex Binary Project Files	⚠️ Raw Text stdout/stderr	⚠️ Text Transcripts Only
10. Coordinated Security SLA (`INV-SLA-10`)	✅ 48h Response / 5d Triage	⚠️ Enterprise Support Queue	⚠️ Quarterly Commercial Updates	➖ Unmaintained Scripts	⚠️ Standard Cloud SLA

Detailed marketing analysis, search keywords, and discovery log: MARKETING-LOG.txt.

6. The 8 Production Usecases

#	Input	Speakers	Output	Typical Workflow
1	Audio	1	Audio podcast, cut	Single-speaker speech cleaning and pause trimming
2	Audio	multiple	Audio podcast, speaker-separated	Diarized multi-track dialogue cleaning
3	Video (A+V)	1	Video cut + animations	Talking-head cut with automated Hyperframes lower-thirds
4	Video (A+V)	multiple	Video + animations + speaker tracking	Multi-speaker interview video with dynamic speaker cards
5	Video → audio only	1/multiple	Audio podcast (video discarded)	Extraction of pristine audio podcast from video footage
6	Audio	1/multiple	Fully generated explainer video	Voice track with AI-orchestrated motion graphics visuals
7	Audio	1	Audio + animated cover	Podcast track wrapped in an animated vinyl/waveform loop
8	Audio/brief	1	Ad clip (15–60 s, 16:9 + 9:16)	Fast short-form clip factory via Hyperframes

For complete step-by-step guides for each mode, refer to docs/USECASES.md.

7. Getting Started & Setup

Create config: Copy config/settings.example.json → config/settings.json and configure local compute options (local/mac, engines, paths.*).
Tools Directory (<TOOLS_ROOT>): paths.tools_root in your settings.json specifies the root for heavy tools (video-use, ffmpeg, Node ≥ 22). Do not locate this inside a cloud-synchronized folder.
External Prerequisites:
- Local: ffmpeg, Node.js >= 22 (for Hyperframes), Python 3.10–3.13 venv.
- Optional Remote Host: faster-whisper + WhisperX on a remote machine reachable via SSH.
- HuggingFace Token: Only required when engines.multi_speaker is set to whisperx.
Optional: install as a package. pip install -e . (or -e ".[storyboard]" to also pull in clip-storyboard-director) registers the ai-media-editor console script, equivalent to running python editor.py below.

8. CLI Reference & Commands

# Set Python venv path
VENV="<TOOLS_ROOT>/.venv/Scripts/python.exe"

# 1. Environment & toolchain verification
PYTHONIOENCODING=utf-8 "$VENV" editor.py doctor

# 2. Display supported operational use cases
PYTHONIOENCODING=utf-8 "$VENV" editor.py modes

# 3. Prepare a media project (deterministic transcription + take packing)
PYTHONIOENCODING=utf-8 "$VENV" editor.py prepare "/path/to/recording.mp4" --mode 3 --project my-video

# 4. Generate timestamped frame contact-sheets for visual LLM inspection
PYTHONIOENCODING=utf-8 "$VENV" editor.py frames my-video --contact-sheet

# 5. Extract detailed frame interval
PYTHONIOENCODING=utf-8 "$VENV" editor.py frames my-video --from 30 --to 45 --step 0.25

9. Motion Graphics & Music Synthesis

Offline Video-Synced Score (`compose_music.py`)

ai_media_editor.tools.compose_music composes background music following a video storyline — 100% local with zero cloud services. Input is a storyline JSON defining sections with exact time windows, emotions, and intensity levels:

# Compose music from storyline JSON
python -m ai_media_editor.tools.compose_music docs/examples/storyline-roshambo.json -o projects/<name>/assets/score

# Generate storyline template
python -m ai_media_editor.tools.compose_music --init

# Run deterministic synthesis selftest
python -m ai_media_editor.tools.compose_music --selftest

Output: Stereo WAV + MP3 + <name>.notes.json + <name>.mid (Standard MIDI File Type 1).

MIDI Export & High-Fidelity Rendering

The MIDI export enables lossless rendering through professional soundfonts:

Path A — SoundFont (local, free): FluidSynth (winget install FluidSynth) + GeneralUser GS / MuseScore_General: fluidsynth -ni soundfont.sf2 out.mid -F out.wav -r 44100
Path B — Orchestral Libraries: VSCO 2 Community Edition or Salamander Grand Piano.

10. Governance & Runtime Invariants

The architecture enforces 10 strict runtime guarantees across all operating modes:

Invariant ID	Operational Domain	Guarantee & Enforcement Rule	Verification Mechanism
`INV-LOCAL-01`	Privacy & Data Egress	100% Local-First Execution: Raw media, transcripts, and embeddings never leave localhost. Zero cloud calls by default.	`tests/test_core.py`, no cloud sockets in core pipeline
`INV-RUNAS-02`	Privilege & Sandbox	Unprivileged User-Mode (`RunAsInvoker`): Zero administrative or elevated privileges required.	Standard user execution across all scripts
`INV-PREV-03`	Storage & Immutability	Preview-Safe & Non-Destructive Source Media: Input files are strictly read-only; all artifacts write to `projects/<name>/`.	Isolation checks in `tests/test_core.py`
`INV-DETERM-04`	Reproducibility	Deterministic Synthesis & Schema Stability: Fixed seed for procedural audio and rigid adherence to Scribe JSON schema.	`tests/test_compose_music.py`, `ai_media_editor/stt/scribe_schema.py`
`INV-BOUNDARY-05`	File Traversal Safety	Anti-Traversal & Project Containment: Filenames and project stems cannot escape the project boundaries.	`test_project_names_cannot_escape_projects`
`INV-SUBPROC-06`	Process Sanitation	Clean Lifecycle & Window Suppression: Preloaded wrappers suppress console window flashing on Windows.	`ai_media_editor/tools/hf.cmd`, `ai_media_editor/tools/hide-windows.cjs`
`INV-PARITY-07`	Multi-OS Support	Cross-Platform OS Parity: Identical execution and path handling across Windows, Linux, and macOS.	Multi-OS GitHub Actions CI matrix
`INV-SYNC-08`	Concurrency & Locks	Cloud-Sync & Multi-Agent Lock Discipline: Heavy tools and projects excluded from cloud sync conflicts.	`.gitignore` conflict & lock filter rules
`INV-DOCS-09`	Accessibility & Discovery	Multimodal LLM Readiness & Bilingual Parity: Frame contact sheets for vision LLMs; full EN/DE documentation parity.	`llms.txt`, `README.md`, `README_de.md` contract tests
`INV-SLA-10`	Security & Incident Response	48h Security Response & 5-Day Triage SLA: Formal vulnerability reporting via GitHub Advisories and maintainer emails.	`SECURITY.md`, `test_security_policy_bilingual_parity`

11. Security & Privacy SLA

Local-First Privacy: Transcription, frame extraction, and cut calculation run entirely offline on local hardware.
Sensitive Media Containment: Source recordings and project output reside strictly within projects/ (gitignored).
Vulnerability Response Commitment:
- Initial Response SLA: Within 48 hours for confirmation of submitted vulnerability reports.
- Technical Triage SLA: Within 5 business days with severity assessment.
- Reporting Channel: GitHub Security Advisories or direct email to security@open-bricks.org, security@ellmos.ai, support@lukasgeiger.com, and lukas@open-bricks.org.
- For full details, see SECURITY.md.

12. Sibling Projects & Ecosystem Matrix

ai-media-editor operates as a specialized multimedia orchestration engine within the open-bricks and ellmos-ai ecosystem:

Repository	Organization	Domain / Purpose	Ecosystem Interoperability
`clip-storyboard-director`	`ellmos-ai`	Generative storyboard and scene director	Scene sequence generation for video usecases
`assistant-core`	`ellmos-ai`	Conversational supervision & agent runtime	Orchestrating autonomous media editing tasks
`decision-clicker`	`ellmos-ai`	Interactive user review gateway	Reviewing candidate cuts and edit decisions
`lock-master`	`ellmos-ai`	Fail-closed multi-agent locking framework	Protecting concurrent project files and render locks
`clutch`	`ellmos-ai`	Process supervisor and task scheduler	Supervising long-running rendering & STT jobs
`system-explorer`	`ellmos-ai`	Local hardware & capability discovery	Detecting GPU, CUDA, and hardware acceleration
`roblox-studio-core`	`ellmos-ai`	Headless 3D studio capture and automation	3D visual assets and animation frames
`usb-podcast-studio`	`entertain-and-more`	USB audio hardware capture & broadcast	High-fidelity recording source for podcast modes
`BattleStage`	`entertain-and-more`	Server-authoritative physics simulation	Game footage and replay video processing
`DevCenter`	`dev-bricks`	Developer environment & workspace manager	Managing local tools and development venvs
`MethodenAnalyser`	`dev-bricks`	Static Python code metrics & method audit	Code quality analysis of media editor modules
`ExplorerPro`	`file-bricks`	High-speed desktop file manager (PySide6)	Visual project browsing and media file organization
`ProFiler`	`file-bricks`	Deep directory analysis and inspection	Media assets indexing and cache analysis
`CloudLockFixer`	`file-bricks`	Multi-host synchronization conflict fixer	Cleaning lock files and cloud conflict copies
`FormularErstellen`	`doc-bricks`	Dynamic PDF & document layout generator	Generating project reports and production summaries
`open-bricks`	`open-bricks`	Umbrella open-source standards organization	Canonical ecosystem governance & license parity

13. Quality Gates & Testing

Run fast quality checks locally without external STT models or heavy media files:

# Run complete test suite
python -m pytest -ra -v

# Run lint checks
ruff check .

# Validate bytecode compilation
python -m compileall -q .

# Verify CLI modes
python editor.py modes

Windows Console Window Suppression

Hyperframes and Node subprocesses run through ai_media_editor/tools/hf.cmd and ai_media_editor/tools/hide-windows.cjs to enforce windowsHide: true, eliminating distracting console window flashing during rendering. Background details: docs/WINDOWS-KONSOLENFENSTER.md.

14. Machine-Readable Context (`llms.txt`)

LLM crawlers, code assistants, and automated indexing agents can parse repository context directly via llms.txt.

Key Search Phrases:

ellmos-ai/ai-media-editor
local AI media editor video podcast transcription
agent driven video editor with local transcription
Claude Code video podcast editor Hyperframes
faster-whisper WhisperX Scribe JSON video-use
transcript based video cutting local first
Hyperframes motion graphics podcast editor
offline procedural music synthesis storyline numpy
zero egress video transcription and cutting

15. Repository Structure

ai-media-editor/                  (code/docs/projects)
├── CLAUDE.md                     ← agent guide (editor workflow, German)
├── README.md                     ← English overview & 18-point visual architecture
├── README_de.md                  ← German documentation (1:1 anchor parity)
├── editor.py                     ← compatibility shim for documented direct calls
├── ai_media_editor/editor.py     ← orchestrator (prepare / frames / modes / doctor)
├── ai_media_editor/tools/
│   ├── cut_view.py               ← pauses as explicit cut candidates
│   ├── frame_view.py             ← video → timestamped frames ("video-scatterer", UC3/4/8)
│   ├── compose_cover.py          ← UC7: loop a cover over audio
│   ├── compose_music.py          ← storyline JSON → video-synced score (NumPy waveform synthesis)
│   ├── hf.cmd / hf-hidden.vbs    ← Windows console window suppression wrappers
│   └── hide-windows.cjs          ← Node.js preload script enforcing windowsHide: true
├── ai_media_editor/stt/
│   ├── scribe_schema.py          ← Scribe-JSON format (contract with video-use)
│   ├── transcribe_local.py       ← faster-whisper + WhisperX → Scribe-JSON
│   ├── diarize_llm.py            ← token-free text-based speaker assignment
│   └── mac_remote.py             ← compute routing (remote primary, local fallback)
├── config/settings.example.json  ← template: compute, engines, models, paths, HF token
├── brand/design-tokens.css       ← branding tokens for generated animations
├── docs/USECASES.md              ← step-by-step per mode
├── production/                   ← optional generative workflows (cloud gates apply)
├── tests/
│   ├── test_core.py              ← dependency-free regression suite
│   ├── test_compose_music.py     ← storyline & waveform synthesis tests
│   └── test_metadata.py          ← automated contract tests for manifests & discoverability
├── SECURITY.md                   ← private vulnerability reporting and security scope
├── THIRD_PARTY_LICENSES.md       ← Level 1 SBOM and third-party license inventory
├── MARKETING-LOG.txt             ← discoverability, personas, matrix & architecture audit
└── projects/<name>/edit/         ← per project: transcripts/, takes_packed.md, … (gitignored)

<TOOLS_ROOT>/                     (NOT a cloud folder — venv/tools)
├── .venv/                        ← Python venv (faster-whisper, video-use, …)
└── video-use/                    ← cloned browser-use/video-use (unpatched)

16. Changelog & Releases

See CHANGELOG.md for full version history.

Version 0.2.3 (2026-09-20): Pfad B visual architecture & discoverability overhaul, 18-point dual-navigation parity, 10-dimension comparative matrix vs 4 alternatives, 4 formalized target personas ([PERSONA-01] to [PERSONA-04]), Level 1 SBOM Invariant Cross-Reference Table in THIRD_PARTY_LICENSES.md, RunAsInvoker non-elevation certification, and § 521 BGB statutory notice.
Version 0.2.2 (2026-09-12): Pfad A technical hygiene pass, CI runner timeout guardrails (timeout-minutes: 15), automated stale issues/PRs lifecycle management, hardened .gitignore, and optional dependency groups in pyproject.toml.
Version 0.2.1 (2026-09-11): Discoverability pass, dedicated third-party license audit, PEP 621 extended URLs, and initial contract test suite.
Version 0.2.0 (2026-09-09): Hardened local-first pipeline, automated contract tests, multi-OS CI matrix, bilingual parity, dual-mermaid diagrams, and 10 governance invariants.

17. Level 1 SBOM & Third-Party Licenses

ai-media-editor is engineered with an uncompromising commitment to open-source transparency, non-elevation, and offline reproducibility:

100% Permissive Open-Source: All integrated libraries, engines, and runtimes are licensed under permissive open-source licenses (MIT, Apache-2.0, BSD, PSFL) or dynamically linked media utilities (FFmpeg LGPL). There are zero proprietary runtime locks or closed telemetry spyware.
Zero Runtime Cloud Dependencies (INV-LOCAL-01): All transcription, frame extraction, pause detection, and procedural music synthesis execute 100% offline on your hardware.
Unprivileged User Execution (INV-RUNAS-02): All components execute strictly under standard unprivileged user accounts (RunAsInvoker) with zero administrative or root elevation required.
Level 1 SBOM & Governance Mapping: Upstream sources, maintainers, license texts, and governing invariants are cataloged in THIRD_PARTY_LICENSES.md.

18. Statutory Notice & Liability Limitation

This software is provided free of charge under the MIT License as open-source software. Under German statutory law (§ 521 BGB - Gefälligkeitsrecht / gratuitous contracts), liability in the case of gratuitous provision of software is limited to intent (Vorsatz) and gross negligence (grobe Fahrlässigkeit). In particular, no warranties are provided for fitness for a particular purpose, merchantability, or absence of defects.

19. Sister Project: `clip-storyboard-director`

clip-storyboard-director is the Director to this editor's Cutter in a two-agent media production model: it handles pre-production (script breakdown, shot timing, AI video-generator prompting via CDP browser automation, 4D persistence/continuity buffers) and hands off an assembled rough cut for ai-media-editor to finish (non-linear editing, color grading, transitions, audio ducking, final render). Each tool is standalone and does not require the other to function.

It is declared as the storyboard optional dependency in this project's pyproject.toml, pinned to a specific reviewed commit for reproducibility. Install both together with the ai-media-editor console script in one step:

pip install -e ".[storyboard]"
ai-media-editor modes
clip-director --version

clip-storyboard-director is MIT-licensed, same as this project; see its own THIRD_PARTY_LICENSES.md for its component inventory.

Navigation Menu

Repository navigation

About

Repository files navigation

ai-media-editor — local AI media editor (Video · Audio · Podcast)

Quick Navigation

1. Key Capabilities

2. Visual Architecture Flowchart

3. End-to-End Execution Lifecycle Sequence

4. Target Personas & Discoverability

[PERSONA-01] Autonomous AI Coding Agent Developers & Swarm Architects

[PERSONA-02] Local-First Podcasters & Content Creators

[PERSONA-03] AI Video & Motion Graphics Engineers

[PERSONA-04] Enterprise Media Security & Compliance Officers

5. Comparative Matrix vs. Alternatives

6. The 8 Production Usecases

7. Getting Started & Setup

8. CLI Reference & Commands

9. Motion Graphics & Music Synthesis

Offline Video-Synced Score (`compose_music.py`)

MIDI Export & High-Fidelity Rendering

10. Governance & Runtime Invariants

11. Security & Privacy SLA

12. Sibling Projects & Ecosystem Matrix

13. Quality Gates & Testing

Windows Console Window Suppression

14. Machine-Readable Context (`llms.txt`)

15. Repository Structure

16. Changelog & Releases

17. Level 1 SBOM & Third-Party Licenses

18. Statutory Notice & Liability Limitation

19. Sister Project: `clip-storyboard-director`

About

Releases

Packages

Contributors

Languages

Name		Name	Last commit message	Last commit date
.github		.github
ai_media_editor		ai_media_editor
assets		assets
brand		brand
config		config
docs		docs
production		production
projects		projects
tests		tests
.gitignore		.gitignore
AGENTS.md		AGENTS.md
CHANGELOG.md		CHANGELOG.md
CLAUDE.md		CLAUDE.md
CONTRIBUTING.md		CONTRIBUTING.md
DONE.md		DONE.md
LICENSE		LICENSE
MARKETING-LOG.txt		MARKETING-LOG.txt
NOTICE		NOTICE
README.md		README.md
README_de.md		README_de.md
RELEASE_GATE.md		RELEASE_GATE.md
SECURITY.md		SECURITY.md
THIRD_PARTY_LICENSES.md		THIRD_PARTY_LICENSES.md
THIRD_PARTY_LICENSES.txt		THIRD_PARTY_LICENSES.txt
TODO.md		TODO.md
VERSION		VERSION
editor.py		editor.py
ellmos-module.v2.json		ellmos-module.v2.json
llms.txt		llms.txt
pyproject.toml		pyproject.toml
requirements-dev.txt		requirements-dev.txt

Latest commit

History

Folders and files

Repository files navigation

ai-media-editor — local AI media editor (Video · Audio · Podcast)

Quick Navigation

1. Key Capabilities

2. Visual Architecture Flowchart

3. End-to-End Execution Lifecycle Sequence

4. Target Personas & Discoverability

[PERSONA-01] Autonomous AI Coding Agent Developers & Swarm Architects

[PERSONA-02] Local-First Podcasters & Content Creators

[PERSONA-03] AI Video & Motion Graphics Engineers

[PERSONA-04] Enterprise Media Security & Compliance Officers

5. Comparative Matrix vs. Alternatives

6. The 8 Production Usecases

7. Getting Started & Setup

8. CLI Reference & Commands

9. Motion Graphics & Music Synthesis

Offline Video-Synced Score (compose_music.py)

MIDI Export & High-Fidelity Rendering

10. Governance & Runtime Invariants

11. Security & Privacy SLA

12. Sibling Projects & Ecosystem Matrix

13. Quality Gates & Testing

Windows Console Window Suppression

14. Machine-Readable Context (llms.txt)

15. Repository Structure

16. Changelog & Releases

17. Level 1 SBOM & Third-Party Licenses

18. Statutory Notice & Liability Limitation

19. Sister Project: clip-storyboard-director

About

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages

Offline Video-Synced Score (`compose_music.py`)

14. Machine-Readable Context (`llms.txt`)

19. Sister Project: `clip-storyboard-director`