Speech to text

Siv3D/include/Siv3D/OpenAI/Speech.hpp

@@ -19,20 +19,68 @@ namespace s3d
 	{
 		namespace Speech
 		{
+			namespace Model
+			{
+				/// @brief 音声合成モデル tts-1 | Speech Synthesis Model tts-1
+				/// @see https://platform.openai.com/docs/models/tts
+				inline constexpr StringView TTS1{ U"tts-1" };
+
+				/// @brief 音声合成モデル tts-1-hd | Speech Synthesis Model tts-1-hd
+				/// @see https://platform.openai.com/docs/models/tts
+				/// @remark tts-1 よりも高品質な音声を生成します。 | Generates higher quality audio than tts-1.
+				inline constexpr StringView TTS1HD{ U"tts-1-hd" };
+			}
+
+			namespace Voice
+			{
+				inline constexpr StringView Alloy{ U"alloy" };
+
+				inline constexpr StringView Echo{ U"echo" };
+
+				inline constexpr StringView Fable{ U"fable" };
+
+				inline constexpr StringView Onyx{ U"onyx" };
+
+				inline constexpr StringView Nova{ U"nova" };
+
+				inline constexpr StringView Shimmer{ U"shimmer" };
+			}
+
+			namespace ResponseFormat
+			{
+				inline constexpr StringView MP3{ U"mp3" };
+
+				inline constexpr StringView Opus{ U"opus" };
+
+				inline constexpr StringView AAC{ U"aac" };
+
+				inline constexpr StringView FLAC{ U"flac" };
+			}
+
+			inline constexpr size_t MaxInputLength = 4096;
+
+			inline constexpr double MinSpeed = 0.25;
+
+			inline constexpr double DefaultSpeed = 1.0;
+
+			inline constexpr double MaxSpeed = 4.0;
+
 			struct Request
 			{
-				String model = U"tts-1";
+				String model{ OpenAI::Speech::Model::TTS1 };
 
 				String input;
 
-				String voice = U"alloy";
+				String voice{ OpenAI::Speech::Voice::Alloy };
 
-				String responseFormat = U"mp3";
+				String responseFormat{ OpenAI::Speech::ResponseFormat::MP3 };
 
-				double speed = 1.0;
+				double speed = OpenAI::Speech::DefaultSpeed;
 			};
 
-			bool Create(const Request& request, FilePathView path);
+			bool Create(StringView apiKey, const Request& request, FilePathView path);
+
+			AsyncTask<bool> CreateAsync(StringView apiKey, const Request& request, FilePathView path);
 		}
 	}
 }

Siv3D/src/Siv3D/OpenAI/OpenAICommon.hpp

@@ -11,6 +11,7 @@
 
 # pragma once
 # include <Siv3D/String.hpp>
+# include <Siv3D/Duration.hpp>
 # include <Siv3D/HashTable.hpp>
 
 namespace s3d
@@ -20,5 +21,8 @@ namespace s3d
 		// HTTP ヘッダーを作成する
 		[[nodiscard]]
 		HashTable<String, String> MakeHeaders(StringView apiKey);
+
+		// 非同期タスクのポーリング間隔
+		constexpr Milliseconds TaskPollingInterval{ 5 };
 	}
 }

Siv3D/src/Siv3D/OpenAI/SivOpenAIChat.cpp

@@ -21,7 +21,7 @@ namespace s3d
 	namespace detail
 	{
 		// OpenAI のチャット API の URL
-		constexpr URLView ChatV1URL = U"https://api.openai.com/v1/chat/completions";
+		constexpr URLView ChatCompletionsEndpoint = U"https://api.openai.com/v1/chat/completions";
 
 		// チャット API に送信するリクエストを作成する
 		[[nodiscard]]
@@ -77,7 +77,7 @@ namespace s3d
 
 				MemoryWriter memoryWriter;
 
-				if (const auto response = SimpleHTTP::Post(detail::ChatV1URL, headers, data.data(), data.size(), memoryWriter))
+				if (const auto response = SimpleHTTP::Post(detail::ChatCompletionsEndpoint, headers, data.data(), data.size(), memoryWriter))
 				{
 					if (const HTTPStatusCode statusCode = response.getStatusCode();
 						statusCode == HTTPStatusCode::OK)
@@ -120,7 +120,7 @@ namespace s3d
 
 				const auto headers = detail::MakeHeaders(apiKey);
 
-				return SimpleHTTP::PostAsync(detail::ChatV1URL, headers, data.data(), data.size());
+				return SimpleHTTP::PostAsync(detail::ChatCompletionsEndpoint, headers, data.data(), data.size());
 			}
 
 			String GetContent(const JSON& response)

Siv3D/src/Siv3D/OpenAI/SivOpenAIEmbedding.cpp

@@ -22,7 +22,7 @@ namespace s3d
 	namespace detail
 	{
 		// OpenAI の Embeddings API の URL
-		constexpr URLView EmbeddingsV1URL = U"https://api.openai.com/v1/embeddings";
+		constexpr URLView EmbeddingsEndpoint = U"https://api.openai.com/v1/embeddings";
 
 		// Embeddings API に送信するリクエストを作成する
 		[[nodiscard]]
@@ -113,7 +113,7 @@ namespace s3d
 
 				MemoryWriter memoryWriter;
 
-				if (const auto response = SimpleHTTP::Post(detail::EmbeddingsV1URL, headers, data.data(), data.size(), memoryWriter))
+				if (const auto response = SimpleHTTP::Post(detail::EmbeddingsEndpoint, headers, data.data(), data.size(), memoryWriter))
 				{
 					if (const HTTPStatusCode statusCode = response.getStatusCode();
 						statusCode == HTTPStatusCode::OK)
@@ -151,7 +151,7 @@ namespace s3d
 
 				const auto headers = detail::MakeHeaders(apiKey);
 
-				return SimpleHTTP::PostAsync(detail::EmbeddingsV1URL, headers, data.data(), data.size());
+				return SimpleHTTP::PostAsync(detail::EmbeddingsEndpoint, headers, data.data(), data.size());
 			}
 
 			Array<float> GetVector(const JSON& response)

Siv3D/src/Siv3D/OpenAI/SivOpenAIImage.cpp

@@ -11,7 +11,6 @@
 
 # include <Siv3D/OpenAI.hpp>
 # include <Siv3D/JSON.hpp>
-# include <Siv3D/Duration.hpp>
 # include <Siv3D/System.hpp>
 # include <Siv3D/SimpleHTTP.hpp>
 # include <Siv3D/MemoryViewReader.hpp>
@@ -22,10 +21,7 @@ namespace s3d
 	namespace detail
 	{
 		// OpenAI の画像生成 API の URL
-		constexpr URLView ImageGenerationV1URL = U"https://api.openai.com/v1/images/generations";
-
-		// 非同期タスクのポーリング間隔
-		constexpr Milliseconds TaskPollingInterval{ 5 };
+		constexpr URLView ImageGenerationsEndpoint = U"https://api.openai.com/v1/images/generations";
 
 		// 画像生成 API に送信するリクエストを作成する
 		[[nodiscard]]
@@ -59,7 +55,7 @@ namespace s3d
 			const auto headers = detail::MakeHeaders(apiKey);
 
 			// 画像生成 API に非同期でリクエストを送信する
-			AsyncHTTPTask task1 = SimpleHTTP::PostAsync(detail::ImageGenerationV1URL, headers, data.data(), data.size());
+			AsyncHTTPTask task1 = SimpleHTTP::PostAsync(detail::ImageGenerationsEndpoint, headers, data.data(), data.size());
 
 			// タスクが完了するまでポーリングする
 			while (not task1.isReady())

Siv3D/src/Siv3D/OpenAI/SivOpenAISpeech.cpp

@@ -10,8 +10,70 @@
 //-----------------------------------------------
 
 # include <Siv3D/OpenAI.hpp>
+# include <Siv3D/JSON.hpp>
+# include <Siv3D/System.hpp>
+# include "OpenAICommon.hpp"
 
 namespace s3d
 {
+	namespace detail
+	{
+		constexpr URLView TextToSpeechEndpoint = U"https://api.openai.com/v1/audio/speech";
 
+		[[nodiscard]]
+		static std::string MakeSpeechRequest(const OpenAI::Speech::Request& request)
+		{
+			JSON json;
+			json[U"model"] = request.model;
+			json[U"input"] = request.input;
+			json[U"voice"] = request.voice;
+			json[U"response_format"] = request.responseFormat;
+			json[U"speed"] = request.speed;
+			return json.formatUTF8();
+		}
+
+		static bool CreateSpeechImpl(const String apiKey, const OpenAI::Speech::Request request, const FilePath path)
+		{
+			// API キーが空の文字列である場合は失敗
+			if (apiKey.isEmpty())
+			{
+				return false;
+			}
+
+			const auto headers = MakeHeaders(apiKey);
+
+			const std::string data = MakeSpeechRequest(request);
+
+			AsyncHTTPTask task = SimpleHTTP::PostAsync(TextToSpeechEndpoint, headers, data.data(), data.size(), path);
+
+			// タスクが完了するまでポーリングする
+			while (not task.isReady())
+			{
+				System::Sleep(TaskPollingInterval);
+			}
+
+			if (not task.getResponse().isOK())
+			{
+				return false;
+			}
+
+			return true;
+		}
+	}
+
+	namespace OpenAI
+	{
+		namespace Speech
+		{
+			bool Create(const StringView apiKey, const Request& request, const FilePathView path)
+			{
+				return detail::CreateSpeechImpl(String{ apiKey }, request, FilePath{ path });
+			}
+
+			AsyncTask<bool> CreateAsync(const StringView apiKey, const Request& request, const FilePathView path)
+			{
+				return Async(detail::CreateSpeechImpl, String{ apiKey }, request, FilePath{ path });
+			}
+		}
+	}
 }