开源语音转文本和文本转语音，已测试

1 分钟的语音数据也可以用来训练一个好的 TTS 模型（少样本语音克隆）

测试发现：GPT-SoVITS API server starts and responds to TTS inference requests: HTTP 200 with a valid synthesized WAV audio file returned for English text-to-speech using a default reference audio on CPU. 23 分钟。

WhisperX：带词级时间戳的自动语音识别（及说话人分离）

测试发现：WhisperX 3.8.7rc1 is installed in a Python 3.12 venv at /tmp/whisperx-venv with CPU-only PyTorch. The test suite passes (12/12 tests) and the CLI prints help and version correctly. All imports resolve without errors. 3 分钟。

Leon 是你的开源个人助理。

测试发现：Leon server installs, builds, serves HTTP on port 5366, and passes all 371 unit tests with Node.js v24 in CI mode (no local AI or voice models). 10 分钟。

使用下一代 Kaldi 与 onnxruntime 实现语音转文本、文本转语音、说话人分离、语音增强、声源分离和 VAD，无需互联网连接。支持嵌入式系统、Android、iOS、HarmonyOS、Raspberry Pi、RISC-V、RK NPU、Axera NPU、Ascend NPU、x86_64...

测试发现：The sherpa-onnx 1.13.8 C++ source built successfully, all 29 CTest tests passed, the published Python package installed, and both Python and native CLI paths performed real Whisper ASR and generated decodable Piper TTS audio. 7 分钟。

一套用代码为 Claude Code 制作的电影风格工具箱：66 种现成风格，也可以制作你自己的风格。每种风格都是一个风格提示词，加上一部完全由 Claude Opus 5.5 用代码制作的演示影片，不使用视频模型。作者：Lemomo。 · 一套用代码做片的风格工具箱：66 种现成风格，也能做出你自己的

测试发现：With Node 20.20.2, the core, voice, and music installs exit 0, and the real pipeline produced a 6.000-second 1920x1080 24 fps film with 144 H.264 frames and stereo AAC audio that fully decoded with exit 0. 7 分钟。

语音转文字的听写应用，支持本地（Nvidia Parakeet/Whisper）和云端模型（BYOK）。隐私优先，并支持跨平台。

测试发现：npm install succeeds, all 944 tests pass with Node 24 and the custom loader bootstraps, test script runs all test suites without failures. 71 分钟。

一个稳健、高效、低延迟的语音转文本库，具备先进的语音活动检测、唤醒词激活和即时转录功能。

测试发现：In a fresh venv with RealtimeSTT[server,faster-whisper] installed, all 591 unit tests pass, the Whisper golden test transcribes the fixture correctly, and the packaged production server answered 200 on /health, /v1/ready and /v1/capabilities and returned the correct transcript on POST /transcribe-pcm16. 9 分钟。

用于 Python 的语音识别模块，支持多种引擎和 API，在线和离线。

测试发现：SpeechRecognition 3.17.0 installed in a venv; 41 of 56 tests pass with 15 expected skips; Google Speech API recognition returns transcriptions from local audio files; CLI sprc binary prints help. 4 分钟。

一个基于 Apple 的 MLX 框架构建的文本转语音（TTS）、语音转文本（STT）和语音转语音（STS）库，在 Apple Silicon 上提供高效的语音分析。

测试发现：A fresh editable install exited 0, the API server returned HTTP 200 on / and /v1/models, a real Kokoro request produced a decodable 2.8-second 24 kHz mono WAV, and the combined core and TTS test suites passed 1506 tests with 45 skipped. 39 分钟。

您网站的语音识别

测试发现：All 165 tests pass, build produces ESM/CJS/IIFE bundles in dist/, and the library is usable via npm import or script tag. 11 分钟。

基于 FunASR 的视频转录、字幕生成和 LLM 辅助剪辑工具，带有本地 Gradio UI。

测试发现：FunClip v2.2.1 is fully installed in a Python 3.12 venv with all dependencies, all 90 tests pass, the CLI recognition+clipping pipeline processes a real video with Paraformer-Large ASR and produces valid clipped MP4 output, and the Gradio web interface launches and serves HTTP 200 on localhost. 49 分钟。

从 EPUB、PDF 和文本生成带有同步字幕的有声书。

测试发现：All 1592 tests pass, Flask web UI serves HTTP 200 on GET /, and the real Kokoro-82M TTS engine with af_heart voice generates non-silent, ffmpeg-decodable WAV audio from text input. 22 分钟。

开源语音 AI 平台。Vapi 和 Retell 的自托管替代方案。支持本地部署和 BYOK，覆盖语音到语音或 LLM/STT/TTS，带有可视化工作流构建器、原生 MCP 和电话支持。

测试发现：Backend Python venv installed with all core deps, PostgreSQL 16 running on port 5432, Redis 7.4.1 running on port 6379, Node.js v22 deployed, pipecat submodule installed editable, FastAPI app imports with speechmatics as the only blocking provider, 493/514 collectable tests pass against real services. 57 分钟。

开源 AI 视频本地化和配音，用于 YouTube/Bilibili： 语音识别、字幕翻译、声音克隆、音频混合和渲染。 开源 AI 视频翻译配音工具。

测试发现：Backend API is fully operational: uvicorn serves the FastAPI app on port 9998, all 421 backend tests pass, and every major endpoint (health, login, tasks, settings, cookies) responds correctly with expected status codes. 20 分钟。

适用于 mac、windows 和 linux 的本地语音听写

测试发现：Full Rust project compiles on Ubuntu 24.04 x86_64 with no root, all 397 unit tests pass (0 failures), binary runs and responds to --help and --version commands. 28 分钟。

用于实时通信的 python 库

测试发现：The fastrtc library installs in a venv, all 29 tests pass, the Kokoro TTS model generates audio, and a demo app starts under uvicorn and serves its OpenAPI schema. 7 分钟。

一个 100M 参数的多语言 TTS 模型，用于实时 CPU 推理、语音克隆和 48 kHz 立体声生成。

测试发现：Dependencies installed in venv; PyTorch CPU inference via infer.py produces valid WAV audio; FastAPI web server responds to health checks and generates 48kHz stereo speech via POST /api/generate with demo presets. 16 分钟。

OpenAI Whisper 的接近实时实现。

测试发现：WhisperLive is installed in a Python 3.12 venv with all dependencies, 318 of 332 unit tests pass (14 metrics tests skipped as they require a dedicated metrics server port), and the server launches with the REST API health endpoint responding HTTP 200 on port 8000 and the WebSocket server listening on port 9090. 11 分钟。

将文本实时转换为语音

测试发现：realtimetts 0.8.10 installed and built wheel+sdist; 369 of 380 tests pass covering BaseEngine, alignment, language router, sentence splitting, streaming, stream player, release guard (except ssh-keygen), and Qwen studio; Qwen server starts and fails only on CUDA GPU requirement; pyttsx3 SystemEngine requires espeak. 13 分钟。

OpenAI Whisper 自动语音识别网络服务 API

测试发现：The Whisper ASR Box webservice is installed and running: the uvicorn server serves Swagger UI at /docs, returns JSON language detection at /detect-language, and returns transcription text at /asr all with HTTP 200. 14 分钟。

一个集成了 ACE-Step、OmniVoice、Kimi Audio、Piper TTS、GPT-SoVITS、CosyVoice、XTTSv2、DIA、Kokoro、OpenVoice、ParlerTTS、Stable...

测试发现：TTS-WebUI server starts and responds HTTP 200 on both Gradio proxy (7767) and main (7770) interfaces; 102 of 105 tests pass (3 skipped require real API credentials); all core extensions import cleanly; CLI troubleshoot passes. 6 分钟。

[AutoArk] GPA（General Purpose Audio）可以用一个微型模型实现 ASR、TTS 和语音转换。

测试发现：GPA v1.5 native PyTorch ASR, TTS, and training all run on CPU with weights downloaded from HuggingFace. ASR transcribes Chinese and English correctly; TTS produces valid 16 kHz mono audio from text plus a reference clip; training executes a forward/backward step. 10 分钟。

面向 Google Meet、Microsoft Teams 和 Zoom 的开源会议转录 API。自动加入机器人、实时 WebSocket 转录、面向 AI agent 的 MCP 服务器。支持自托管或使用托管 SaaS。

测试发现：11 Python service test suites and the JS terminal client test suite run successfully in a container with Node 22.13 and Python 3.12.3; the compose/Docker deployment path cannot execute without Docker engine being available. 17 分钟。

Python 库和 CLI 工具，用于与 Google Translate 的文本转语音 API 对接。

测试发现：gTTS 2.5.4 installed, all 108 pytest tests pass, and the CLI generates valid MP3 audio via the Google Translate TTS API. 1 分钟。

用于 python 的离线文本转语音合成

测试发现：pyttsx3 v2.99 with eSpeak-NG v1.53.0 backend on Python 3.12.3 (Linux): import, speech synthesis, WAV file saving, property get/set, event loop, and all 9 applicable tests pass. 4 分钟。

面向 16+ 个模型家族的 ggml 语音转文字推理

测试发现：transcribe.cpp builds from source with cmake and GCC 13, produces a working transcribe-cli binary that understands all CLI options and reads 16 kHz mono WAV files, and passes all 39 unit/smoke tests on CPU without a model file. 4 分钟。

开源本地 AI 音频与音乐工作室，支持 macOS 和 Windows。提供 TTS、语音克隆、带歌词的歌曲以及多轨乐谱，100% 离线，zero API 费用。

测试发现：安装、前端构建、Go 核心测试（3 包通过）与 Playwright 浏览器测试（74/74 通过）全部成功，浏览器预览应用服务返回 HTTP 200，独立 CLI（engine v0.7.4）与 audiocpp CPU 引擎（0.7.4）均能启动运行；仅原生桌面 App 因仓库限 macOS/Windows 未在此 Linux 容器构建。. 10 分钟。

GPT-SoVITS ONNX 推理引擎与模型转换器

测试发现：GENIE TTS package installs and builds in a Python 3.12 venv; all 4 language G2P modules produce correct phone IDs; FastAPI server serves all endpoints (/stop, /clear_reference_audio_cache, /load_character, /set_reference_audio, /tts, /unload_character) returning HTTP 200; full TTS inference with predefined character... 9 分钟。

一款出色的浏览器扩展，可一键朗读网页内容

测试发现：All 159 tests pass consistently (0 fail, 0 skipped, 0 todo) across repetitive concurrent execution; the offscreen-manager fixture timeout was relaxed to eliminate a timing-dependent failure. 3 分钟。

用于 Wayland 合成器的语音转文字（支持按键说话）

测试发现：Voxtype v1.1.0 builds and passes its full test suite (1159 passed, 0 failed) on Ubuntu 24.04 amd64 with Rust 1.99.0, after manually providing alsa and clang/LLVM development libraries extracted from apt packages without root privileges. The CLI binary responds correctly to --version and --help. 27 分钟。

主题 → 面向编码 agent 的 4K 旁白视频。v5.3.0：本地 TTS（edge 免费版 + azure，无需外部引擎）、基于清单的 Asset Engine、Remotion 合成、成本控制的 AI 生成、Bilibili/YouTube/小红书/抖音/微信视频号

测试发现：All 255 Python tests pass, TypeScript type-checks pass, and Remotion 4.0.438 renders a 2-frame smoke test to a valid H.264 MP4 verified with ffmpeg decode check; the only source change needed was bumping zod from ^3.23.0 to ^4.3.6 in package.json. 7 分钟。

Soprano：即时超逼真文本转语音

测试发现：Soprano TTS 0.2.0 is installed from source in a venv at /tmp/venv. The CLI produces valid 32 kHz WAV audio on CPU. The FastAPI server serves the /v1/audio/speech endpoint (HTTP 200, valid WAV). The Gradio WebUI module loads successfully. Streaming playback requires the system PortAudio library (not available in this... 15 分钟。

AI Dictation App：开源且本地优先，打字速度快 3 倍，无需键盘。由开源模型驱动，离线可用，快速准确。

测试发现：Installation completed: Node 24.10.0, pnpm 10.34.5, all 1566 packages installed, whisper native addon compiled (.node binary at packages/whisper-wrapper/native/linux-x64/whisper.node), electron 44.3.0 available, workspace builds pass, and the test suite runs (149/150 test files pass, 1505/1508 tests pass, 3... 23 分钟。

扩展 Comfyui 功能的自定义节点

测试发现：The repository installs as a ComfyUI custom_nodes extension: pip dependencies install cleanly, ComfyUI starts on CPU, the server answers HTTP 200, and all 8 AlekPet custom node groups load successfully with zero startup errors. 34 分钟。

基于 Flow Matching 的文本转语音模型，支持表情符号驱动的风格控制

测试发现：Irodori-TTS installs and all core modules import cleanly on CPU. The TextToLatentRFDiT model constructs and runs forward passes. The Gradio web UI builds and serves HTTP 200. Inference requires downloading a HuggingFace checkpoint. 8 分钟。

多后端 Whisper 应用。速度极快，针对 Mac ARM 优化，安装简单。输入本地文件或 URL，该服务将使用 Whisper AI 进行转录。完全私密且免费。

测试发现：transcribe-anything 4.1.2 installed in a uv-managed Python 3.11 venv with Python 3.10 available for iso-envs; 265/278 tests pass, CLI help works, all backends ready for first-use env creation. 21 分钟。

视频同步翻译。视频配音。

测试发现：SoniTranslate web app built with Gradio 4.19.2 launches via python app_rvc.py --cpu_mode and serves a fully constructed UI at http://127.0.0.1:7860 returning HTTP 200 on / and /config endpoints. 22 分钟。

一个用于长篇幅、多说话人对话合成的多语言模型，支持灵活的说话人控制和零样本语音克隆。

测试发现：MOSS-TTSD v1.0 model loads from HuggingFace, processor encodes conversations, forward pass returns logits, and generate() produces audio codes and text tokens on CPU with bfloat16 precision. 25 分钟。

能够实时流式传输对话音频的 TTS 模型

测试发现：Dia2-1B model loads and generates valid 24 kHz mono WAV audio on CPU (bfloat16) via both the Python API and CLI, producing non-silent output with correct speaker-tagged transcription and word timestamps. 12 分钟。

Whisper-Flow 是一个框架，旨在使用 OpenAI 的 Whisper 模型实现音频内容的实时转录。与上传后处理整个文件的“batch mode”不同，Whisper-Flow 接受连续的音频片段流，并立即生成增量转录。

测试发现：Whisper Flow V1.1.0 server launches on port 8181, serves /health (200) and /ready (200) endpoints showing model loaded and 0 active sessions, all 29 core tests pass with 97% coverage, and all core modules import correctly. 18 分钟。

使用自托管 Whisper 和 WebSocket 的近实时音频转录（Python/JS）

测试发现：A fresh venv installs cleanly; the VoiceStreamAI server launches on 127.0.0.1:8765 with the token-free silero VAD plus the real faster-whisper-tiny model and streams a real transcription back over WebSocket ('Good morning, everyone!'), while 2 of 3 unittest tests pass and the pyannote VAD test remains blocked by the... 15 分钟。

使用词错误率 (WER) 等相似度度量来评估您的语音转文本系统。

测试发现：The jiwer package is installed and functional: all 74 tests pass, the CLI prints its help, and the Python API computes WER/MER/WIL/WIP/CER correctly including empty-reference edge cases. 1 分钟。

用于 PHP 的 Vonage REST API 客户端。支持 SMS、Voice、Text-to-Speech、Numbers、Verify (2FA) 等 API。

测试发现：Vonage client-core installs with composer update and its full PHPUnit suite passes 1096/1096 after the Guzzle 8 version-guard fix in src/Client.php, with SMS send verified end-to-end against a local mock HTTP client. 4 分钟。

由你的 AI agent 制作的解说视频和产品演示。免费且开源：本地语音（Kokoro）、单词计时（Whisper）和画布渲染器将脚本转换为带旁白的 MP4。

测试发现：explainroo v0.1.0 is installed with Node.js 20.20.2, Playwright Chromium headless shell, Kokoro TTS, Whisper ASR, and all npm dependencies. The project initializes, voices, checks, stills, sheets, and renders draft MP4s. The how-dns-works example produces a 68s video with 8 scenes. 29 分钟。

C++ ggml 运行时中心，支持多语言 ASR 和 TTS 模型：Cohere Transcribe、Parakeet TDT、Voxtral、Canary 1B v2 等，外加通用强制对齐等功能。

测试发现：CrispASR v0.8.37 builds, runs CLI transcription (Whisper base model), and serves an HTTP API, all tested end-to-end with real audio in this container. 36 分钟。

Claudemy：一个开源的 Udemy 和 Coursera 替代方案。向 Claude 提供一个主题，它会搜索网络并构建一个由 2 分钟音频课程组成的课程，并附有引用来源。一个免费的 Claude Code 技能。

测试发现：All 39 regression tests pass. beautifulsoup4 is installed in a venv, ffmpeg is available, all four CLI scripts (scrape, check_lessons, render_audio, package_class) emit correct usage docs, the lesson checker correctly validates scripts and citations, and the package class script combines cards into a deliverable. 2 分钟。

一个使用 Playwright 解决 reCAPTCHA v2 和 v3 的 Python 库

测试发现：All 16 tests execute without errors against live reCAPTCHA demo pages and antcpt.com; v2 solvers correctly raise RecaptchaRateLimitError against Google's rate limits, v3 solvers solve antcpt.com successfully, and recaptcha_not_found tests raise proper exceptions when no captcha is present. 7 分钟。

AI 头像/数字人平台：上传照片，克隆声音，与任何面孔实时对话并带有唇形同步视频。开源，自托管。Claude · Whisper · Chatterbox · MuseTalk。

测试发现：Backend FastAPI server runs on port 8000 with SQLite storage, 191 pytest tests pass, frontend Next.js builds and serves on port 3000, full REST API flow (register/login/me/avatar upload/session create/guest) works end to end in local-dev mode without Docker, Redis, or third-party API keys. 10 分钟。

在自己的电脑上构建 AI 电话 agent，并使用 AI 呼叫者测试每项更改。开源，无需 API 密钥即可免费运行，使用自己的密钥时每次通话只需几分钱。支持导入 Dialogflow CX。

测试发现：Both packages (vpd-vk 0.3.0, vpengine 0.3.0) installed in .venv-engine; 453 tests pass; engine starts and answers health and chat session endpoints. 23 分钟。

阿拉伯语优先的生成式语音识别：Audar-ASR-V1 (Flash + Turbo)。在 Open Universal Arabic ASR 排行榜上排名 #1。模型卡、benchmarks 与推理。

测试发现：Audar-ASR-V1-Flash model loads and transcribes on CPU. All three reference API functions (transcribe, transcribe_long, stream_transcribe) work correctly. 27 分钟。

Dramaxis (剧枢)，电影级多智能体影视叙事与智能解说平台 (Tauri 2 + Rust + React 19)。6 大智能体协同导演、5 轨毫秒级对齐时间轴、四道质量闸门硬核质检、本地优先沙箱架构与剪映草稿深度互通。

测试发现：All 280 frontend tests pass, frontend production build succeeds (299 modules), 10 of 11 Rust workspace crates compile and all their tests pass (core=13, intake=17, export=18), VitePress docs build succeeds. Tauri desktop shell (root Cargo.toml + update crate) blocked by missing libwebkit2gtk-4.1-dev system packages... 11 分钟。

一句话让 AI 帮你做产品宣传片。把产品名或网址告诉 Cursor、Claude Code、Codex 等 AI 编程工具，它会自己调研产品、画出多种风格样图让你挑选、截取 4K 真实页面、按音乐节拍卡点剪辑、生成 AI 旁白并逐句转写校验、自动审图，最后交付一支 4K...

测试发现：Template project installs, fonts subset correctly, stills render, 1080p preview video renders cleanly, web player builds; lint passes; all doctor checks pass except optional ElevenLabs key. 10 分钟。

本地语音听写和会议记录工具，支持 Windows 和 Linux。按住热键进行听写，或录制包含系统音频的长会议。使用 Whisper 进行转录，自带 LLM 摘要功能。开源。

测试发现：pnpm+uv deps installed, TypeScript+Python code compiles/typechecks, frontend Vite build succeeds, 331/331 backend tests pass, and the Qt+Pyloid desktop app launches on the virtual display and loads the Whisper speech model. 22 分钟。

开源 AI 会议 copilot - 实时转录、回声消除和 AI 辅助。捕获系统音频 + 麦克风，via WebRTC AEC3 消除回声，使用 Deepgram 进行转录，并在会议期间提供 Claude/OpenAI 帮助。可在 macOS 和 Windows 上本地运行。

测试发现：Raven 2.5.0 Electron app builds from source, runs tests (1088 passed), and launches on Linux with Vite dev server and Electron main process. 3 分钟。

使用 NVIDIA Parakeet 在 Rust 中实现极快的语音转文字、说话人分离和流式处理（甚至可在 CPU 上运行）

测试发现：parakeet-rs compiles, all 24 tests pass, and the full ONNX inference pipeline runs successfully with a verified model. 4 分钟。

用于 Node.js 的 Vonage API 客户端。API 支持 SMS、Voice、Text-to-Speech、Numbers、Verify（2FA）等。

测试发现：The @vonage/server-sdk monorepo installs cleanly under Node 22.23.4, all 618 tests pass, all 28 packages build to CJS/ESM/DTS, lint passes, and a real end-to-end call against a local mock returned a parsed account balance of 10.28 EUR and an SMS response with message-count 1 and status 0. 5 分钟。

一个轻量级的 Python 包，用于使用 ONNX 模型进行自动语音识别。

测试发现：All 249 core tests pass, package imports correctly, CLI displays help with all supported model names, and whisper-tiny model executes real ASR inference via Hugging Face. 12 分钟。

在 Obsidian 中使用 Whisper 进行语音转文本

测试发现：All 69 tests pass (5 test files), TypeScript type-checking passes, esbuild bundles main.js successfully with no errors. 1 分钟。

一款桌面应用，可从文件、URL、YouTube 视频或麦克风转录音频和视频，然后编辑、翻译、总结、添加字幕，并与转录内容对话。

测试发现：All 554 tests pass, the desktop GUI opens a mapped window on Xvfb, and the CLI transcribes audio files with WhisperX on CPU. 8 分钟。

AAAI 2025：编解码器确实重要：探索音频语言模型中编解码器的语义缺陷

测试发现：X-Codec model builds, imports, runs encode/decode on real audio across all bandwidths, and produces valid WAV output. 8 分钟。

OpenAI Whisper 模型的 C/C++ 移植

测试发现：whisper.cpp builds from source and transcribes the JFK speech sample correctly with the base.en model on CPU, returning 7 passing tests and a working HTTP server. 5 分钟。

创建和编辑 AI 视频教程，包含带插图的课程、富有表现力的演示者、独特的声音和原生时间线。Windows 桌面应用，支持本地模型和云服务提供商。

测试发现：All 6 workspace packages build successfully. JS unit tests pass 156/156 (packages/contracts/scenes/themes). Renderer tests pass 100/117 (17 pre-existing version-mismatch failures). Python pipeline tests pass 942/956 (2 pre-existing failures: Linux path normalization vs Windows test, and stale manifest byte count). The... 22 分钟。

自托管的 TikTok、Douyin 与 Bilibili 数据采集器和无水印视频下载器，提供异步 REST API、面向 AI agent 的 MCP 服务器、CLI 和 Web 控制台，并支持对已下载视频进行离线语音转文字。具备自愈身份池和 PostgreSQL 归档，一条 docker compose up...

测试发现：All 2960 tests pass (2450 unit + 510 integration), the CLI outputs version 5.0.1 and lists 10 commands, the API server starts and serves healthz/readyz/swagger/console root with 200, the web console SPA is built, and both the PostgreSQL+TimescaleDB and Redis backends and the Go downloader sidecar run. 48 分钟。

一个实时语音运行时，让 Agent 保持对话、工作和在场。面向 AI Agent 的实时语音运行时。

测试发现：The qwen-audio-agent v1.11.0 project installs, builds, launches a Gateway serving HTTP 200 on port 3101, and passes all 1789 automated tests with zero failures. 9 分钟。

Vocello：适用于 Apple Silicon 的私人语音工作室。编写脚本、选择或描述声音，在 Mac 和 iPhone 上本地生成语音。原生 Swift 和 MLX，无需 Python，无需云端。Mac 应用和命令行现已推出，iPhone 测试版在 TestFlight 上。（原名 QwenVoice。）

测试发现：On this Linux container the repository's deterministic contract gate passes cleanly, the Python suite passes 1526 tests with 2 unfixable lsof-dependent tests failing, the website's 19 unit tests and production build pass under portable Node 22, and a vite preview server served the prerendered site and both... 12 分钟。

基于 CTranslate2 的 Whisper 命令行客户端，兼容原始 OpenAI 客户端。

测试发现：Installed in a venv; the CLI answers version 0.5.7, all 17 unit tests pass, and 6 of 7 e2e tests pass including real small/medium-model transcription of the bundled samples matched against reference outputs, with only diarization failing because it needs a real HuggingFace token and pyannote/torch. 22 分钟。

Dramaxis (剧枢)：电影级多智能体影视叙事与智能解说平台 (Tauri 2 + Rust + React 19)。6 大智能体协同导演、5 轨毫秒级对齐时间轴、四道质量闸门硬核质检、本地优先沙箱架构与剪映草稿深度互通。

测试发现：Frontend installs, type-checks, builds, and its full Vitest and Playwright suites pass (157 unit tests, 8 e2e tests), and all 12 non-GUI Rust crates build and pass 79 unit plus 2 doc tests; the Tauri desktop shell still cannot be compiled because its system GTK/WebKit dev libraries are absent from this container. 28 分钟。

ASR/STT 字幕生成器。使用 Qwen3-ASR、本地 LLM、Whisper、TEN-VAD。针对 JAV 的噪声鲁棒。

测试发现：Project installed from source at /work/repo with 79 packages in a Python 3.12 venv. All 4 verification commands pass: import, --version (1.9.3), --help (full CLI), --dump-params (safe config probe). 617/617 clean tests pass; 43 failures and 11 errors are pre-existing (version drift, restructured config, missing... 11 分钟。

面向 Claude Code 的 AI 原生视频制作工具包

测试发现：Hello-world, sprint-review template, render-baseline test, and concept-explainer-short template all render valid MP4 videos; migration script fixed and installs all 27 skills/commands; verify_setup.py confirms all prerequisites met. 16 分钟。

一款出色的浏览器扩展，一键朗读网页内容。

测试发现：Read Aloud extension package builds cleanly (474KB ZIP), all 49 JS files pass node syntax check, the extension loads into Chromium 127 via Playwright without errors, manifest declares all TTS permissions and a service worker, and 14 distinct TTS engine implementations are wired in tts-engines.js. 5 分钟。

用于 ComfyUI 的 OmniVoice TTS 节点：零样本多语言文本转语音，支持语音克隆、语音设计和多说话人对话

测试发现：5 OmniVoice TTS ComfyUI nodes (Longform TTS, Voice Clone, Voice Design, Multi-Speaker, Whisper Loader) installed and registered; OmniVoice-bf16 model loads from HuggingFace cache, generates valid 24kHz audio waveform from text on CPU in ~9 seconds for 2-second utterance. 5 分钟。

开源、本地优先的 Granola/Otter 替代品，Claude Code、Codex、Cursor 及任何 MCP 客户端均可查询。会议、通话和语音备忘录在设备端转录为你拥有的 markdown。

测试发现：The minutes CLI v0.27.0 binary builds and runs: 59 commands available, demo generates 5 sample meetings with correct markdown frontmatter and YAML metadata, capabilities endpoint reports 50+ features, whisper-guard passes 14/14 tests, 1827/1864 minutes-core tests pass with 37 container-limited failures, 3/4... 30 分钟。

一款免费、开源且可扩展的语音转文本应用，完全离线运行。

测试发现：The Handy app builds from source (debug binary + frontend), passes its full Rust unit test suite (271/271) and Playwright frontend tests (2/2), and launches and stays up on the Xvfb virtual display, enumerating its CPU compute device and 69-model catalog via CLI. 67 分钟。

一个实时无声语音识别工具。

测试发现：AVSR model loads on CPU and produces transcriptions from video features; MediaPipe face detector initializes; all Python imports resolve; webcam capture fails in headless environment (no /dev/video0) but app otherwise works. 23 分钟。

适用于 Flutter 的端侧 AI：Gemma 及其他 LLM，支持多模态、函数调用、嵌入、RAG 和语音，可在 Android、iOS、Web、macOS、Windows 和 Linux 上运行。

测试发现：All 14 Dart pub workspace packages with test directories pass their test suites on Linux x64 with Flutter 3.47.0 and Dart 3.13.0, with 0 errors and all 567 analyzer findings at info level only. 11 分钟。

VRCT（VRChat 聊天框翻译与转录）

测试发现：Python backend starts and completes full initialization (Whisper model, CTranslate2, OSC), Vite frontend builds and serves, all 936 pytest tests pass, OCR engine is ready with downloaded models, Rust toolchain installed; Tauri native build blocked by missing webkit2gtk-4.1 system libraries (not installable without... 25 分钟。

免费开源的 AI 桌面应用，用于有声书、播客、纪录片、电影和动画视频，支持本地模型、云 API 和可编辑的故事板。

测试发现：Dependencies installed, 819 non-multimedia tests pass (0 failures), the engine host HTTP server answers 200 on /health, and 4 code/test bugs were fixed; the Qt GUI cannot launch because no PipeWire audio daemon runs in this container. 31 分钟。

Amphion (/æmˈfaɪən/) 是一个用于音频、音乐和语音生成的工具包。其目的是支持可复现研究，并帮助初级研究人员和工程师入门音频、音乐和语音生成的研究与开发。

测试发现：Core dependencies installed, evaluation metrics compute correctly on audio pairs, model classes import cleanly. pyworld-dependent f0 metrics unavailable without python3-dev headers. Cython monotonic_align extension uncompiled for same reason. 25 分钟。

一个基于 Apple 的 MLX 框架构建的文本转语音（TTS）、语音转文本（STT）和语音转语音（STS）库，在 Apple Silicon 上提供高效的语音分析。

测试发现：mlx-audio v0.5.6 installed in a Python venv with mlx[cpu] backend, 354 of 376 tests pass (22 skipped due to missing torch/diffusers for parity), FastAPI server starts and responds 200 on /v1/models, all five CLI tools print help. 8 分钟。

ODS V3 预发布版：在正式 V3 发布前进行公开测试和优化。将你的 PC、Mac 或 Linux 设备变成私有 AI 服务器。

测试发现：Project builds correctly, all 418 bats tests, 14/14 smoke tests (after ai_err fix), ~300 shell/Python contract tests, and the vite frontend build pass. The Docker runtime install cannot complete in this environment, but the dry-run installer, CLI (-v2.6.0), all standalone test suites, and the frontend are functional. 15 分钟。

一个自主、上下文感知的 AI 桌面伴侣，使用 Python 构建，具有实时屏幕视觉、自定义 ONNX 语音合成、活动窗口追踪，以及带有反应式面部表情的动态浮动 UI。

测试发现：All 18 modules initialize correctly: EventBus, KinematicsEngine, WindowManager, FaceRenderer, ChatBubble, InputBox, StateMachine, CognitiveEngine (Groq mode), Vision sensors (semantic + mss), WindowTracker, AutonomyEngine, and VoiceEngine all boot to 'B is alive' with the Qt event loop running at ~62fps. Windows-only... 28 分钟。

一个开源模型系列，用于长语音、对话合成、声音设计、音效和实时流式 TTS

测试发现：MOSS-TTS Python package installs and builds; FastAPI server starts and serves /health endpoint on port 8768; CLI entry point prints usage; all core Python modules import; the project is ready for model weight download and GPU inference but cannot run inference in this container because no CUDA GPU is available and the... 31 分钟。

ElevenLabs API 的官方 Python SDK。

测试发现：The elevenlabs Python SDK v2.69.0 is installed in a venv and all 219 tests pass using httpx.MockTransport to mock ElevenLabs API endpoints. 15 分钟。

本地实时多模态 AI，功能类似 GPT-Live

测试发现：Project installs, builds, and the server starts with Gemma 4 E2B loaded and serving HTTP+WebSocket; the end-to-end AI pipeline (turn detection, llama.cpp inference, transcript parsing, TTS) all initialize and run correctly when tested manually, but the full test suite cannot complete because CPU-only inference exceeds... 77 分钟。

面向 Douyin 直播电商的 Windows 本地 AI Agent Studio，贯通主播发现、直播洞察、直播复盘与短视频内容编导的统一智能工作流。

测试发现：The web UI serves 200 with the LiveAgent Studio title, the local gateway health endpoint verifies both mock DashScope and mock OSS as configured/verified, all 37+16+3 agent unit tests plus the frontend build tests pass, and the full video breakdown pipeline produced and decoded a real timestamped Excel transcript. 27 分钟。

适用于 Windows、Linux 和 macOS 的便携式本地人工智能工作室。零配置图形用户界面，支持图像生成、GGUF LLM、文本转语音和语音转文本。

测试发现：Frontend deps installed, React UI builds and is served on localhost:1420, all 13 regression tests pass, server API endpoints respond correctly. 7 分钟。

Talking Head (3D)：一个用于实时唇形同步的 JavaScript 类，使用全身 3D 虚拟形象。

测试发现：TalkingHead 3D avatar library installed with npm, local brunette.glb avatar loads successfully in headless Chrome via WebGL, and the full test suite of 32 streaming API tests passes with 0 failures. 14 分钟。

适用于 Apple Silicon 的高性能、兼容 OpenAI 和 Anthropic 的 LLM 推理服务器。支持原生 MLX、连续批处理、多模态模型、MCP 工具调用以及 Claude Code。

测试发现：vllm-mlx 0.5.0 installs into a venv on x86 Linux, its CLI help command prints usage for 8 subcommands, and 1559 of its Linux-compatible unit tests pass. 23 分钟。

官方的 MiniMax Model Context Protocol (MCP) 服务器，支持与强大的文本转语音、图像生成和视频生成 API 进行交互。

测试发现：Package installed in a venv, all 9 pre-existing tests pass (3 server + 6 utils), and the MCP server binary launches and waits for stdio input. 2 分钟。

一个全面的 ComfyUI 集成，用于 Microsoft 的 VibeVoice 文本转语音模型，支持在您的 ComfyUI 工作流中直接进行高质量的单人和多人语音合成。

测试发现：5/5 nodes registered. Torch 2.14.1+cu130, Transformers 4.57.6, Diffusers 0.39.0. Needs real model files + Qwen2.5-1.5B tokenizer for TTS. 22 分钟。

OuteTTS 模型的接口

测试发现：outetts 0.4.4 installed with all dependencies including llama-cpp-python 0.3.9 (CPU), Python modules import cleanly, default speakers load, JS test suite 12/12 passes. 15 分钟。

Linux 原生语音转文字：快速、准确、私密且可定制的系统级听写

测试发现：Python venv with all test and runtime dependencies installed; 1146/1146 tests pass; CLI responds correctly for help, config, status, transcribe routing; main service blocked only by missing system libportaudio2 which requires root to install. 17 分钟。

一个语音聊天应用

测试发现：The QuiLLMan frontend and websocket adapter run locally via uvicorn on port 8000 with a mocked Modal layer serving both the React frontend (200 on /) and a FastAPI websocket endpoint (200 on /status, echo responses on /ws). 11 分钟。

Botium 语音处理

测试发现：The Botium Speech Processing frontend server runs on port 56000 serving status, STT, TTS, languages, voices, and Swagger endpoints, with mock backends for MaryTTS and Kaldi substituting for the Docker-dependent default services. 6 分钟。

Lobe TTS：一个高质量且可靠的 TTS/STT 库，适用于服务器和浏览器。

测试发现：Dependencies installed; library builds and type-checks on Node 22.14.0; createEdgeSpeech connects to a WebSocket, sends speech.config and ssml frames, and returns a 200 Response containing the mocked audio payload. 17 分钟。

React Native Expo 项目的语音识别

测试发现：The library builds and type-checks, its example web app serves and renders with no page or console errors, and its web speech module start/stop/event wiring works against a mock SpeechRecognition in Chrome. 53 分钟。

使用 Whisper 和 LLMs（GPT、Claude 等）将语音转录并翻译为 LRC 文件。

测试发现：openlrc 1.7.0a2 is installed and built in a Python 3.12 venv; all 186 mock-based unit tests pass; the httpx/httpx2 compatibility bug in ClaudeBot and GPTBot constructors is fixed. 27 分钟。

测试发现：All 17 examples install and run: every Next.js example builds (pnpm run build exit 0) and its / and API routes answered 200 against schema-correct local mocks returning real decodable MP3 audio, transcripts, and tokens; the CLI TS/Python quickstarts produced output.mp3 (ffmpeg-decode clean) or printed the transcript... 64 分钟。

一个简单的 Azure TTS API 文本转语音客户端。

测试发现：aspeak 6.1.0 builds and runs from source with PKG_CONFIG_PATH and LIBRARY_PATH workarounds; CLI text and SSML synthesis, stdout output, and offline subcommands (list-formats, list-qualities, --help, --version) all work; real Azure endpoint tests skipped due to missing credentials. 10 分钟。

Deepgram 官方 Python SDK。

测试发现：deepgram-sdk 7.10.0 is installed in a Poetry virtualenv and its full test suite passes (1052 passed, 4 skipped) against a local WireMock-compatible mock, with sync and async client calls returning parsed typed responses and ruff plus mypy clean. 7 分钟。

开源 AI 面试助手和会议副驾驶：在 Zoom、Google Meet 和 Teams 通话中实时回答，设备端转录，屏幕共享时隐藏。免费替代 Cluely、Interview Coder 和 Final Round AI，支持 Windows、macOS、Linux。

测试发现：The app installs and builds on the bundled toolchain, 123 unit tests and 44 relay tests pass, the Electron app launches (windows visible on Xvfb), and a chat request travels from the built app through the local relay to the OpenAI-compatible mock upstream and back with HTTP 200. 10 分钟。

一个使用 Microsoft Edge Read Aloud API 的简单 Azure Speech Service 模块。

测试发现：pnpm install and TypeScript build succeed, all 10 Jest tests pass using a local WebSocket mock of the Microsoft Edge Read Aloud API. 26 分钟。

从 AI 工具到使用 FFMPEG、Microsoft Edge 朗读和 OpenAI Whisper 模型创建 TikTok 视频

测试发现：Package installs (uv sync --frozen), CLI help works, list-voices queries real Edge TTS API returning 9 Male + 8 Female English voices, create pipeline runs end-to-end with yt-dlp YouTube download, Whisper transcription, FFmpeg video composition delivering a valid 2.3MB MP4, and pytest passes 1/1. 7 分钟。

Sayna 是一个统一的语音层，用于 AI Agents，能够无缝集成到现有的智能体框架中。

Voicy（@voicybot）：一个 Telegram bot，可将群组和私聊中的语音消息转换为文本。

测试发现：The repo installs and builds (yarn install + tsc), and 32 proof scripts including the worker end-to-end vNext QAs pass against a locally-started mongod; live-Telegram checks were not run because they need a real bot token, and the local-whisper proof was made platform-independent. 6 分钟。

将电脑或麦克风音频实时翻译成字幕。

测试发现：Full Rust library unit-test suite (421 tests) passes, frontend vitest suite (120 tests) passes, TypeScript typecheck and production Vite build succeed, ESLint clean, with all system GTK/WebKit GLib dependencies resolved from apt-downloaded .deb packages extracted to a local prefix. 21 分钟。

面向流式传输和设备端使用的优化 Whisper 模型

测试发现：In this container the Python package installs and imports, the streaming pipeline and examples/server.py run against a local mock ASR endpoint returning word timestamps, and the Electron client launches on the virtual display and completes a session-create/audio-chunk/process transcription round trip; real CUDA/CoreML... 34 分钟。

适用于 Claude Code、Cursor、Codex 和 Gemini CLI 的 24 个跨平台 agent 技能：数据库、消息传递、研究、TTS、DevOps 和 Google Workspace

测试发现：All Python deps install and the full pytest suite passes (11 passed, 1 skipped); Outline, Telegram, Google TTS, ElevenLabs, Imagen, and NotebookLM E2E smoke are verified working against local mocks with production defaults preserved, while real-database skills and credential-gated third-party skills remain unverified... 40 分钟。

用于 VibeVoice TTS 的 ComfyUI 自定义节点。富有表现力、长文本、多说话人的对话音频。

测试发现：ComfyUI-VibeVoice v2.11.0 core modules and nodes install successfully; 1365 of 1634 tests pass under mocked ComfyUI stubs; remaining 230 failures require a real ComfyUI installation with model weights and GPU. 31 分钟。

开源语音AI SDK。面向希望拥有完整技术栈的开发者的 Vapi/Retell 替代方案。用 4 行代码为你的 AI agent 分配一个电话号码，支持 Python 和 TypeScript，采用 MIT 许可证，集成 Twilio、Telnyx 和 Plivo。

测试发现：Python SDK installs and runs 3037 passing tests. TypeScript SDK installs, compiles with zero errors, builds, and all test files pass. Python CLI responds with usage text. Two code bugs (SSRF IPv4-mapped blocking, mtime tiebreaker) and one tooling issue (vitest/vite version mismatch) were diagnosed and fixed. 41 分钟。

来自电脑或麦克风音频的实时翻译字幕。

测试发现：In this container, mimi builds cleanly (all check.sh gates green) and the dev binary launches and stays up on the Xvfb display with settings, overlay, control and tray windows mapped, its UI-test synthetic session toggling between listening and idle via the Ctrl+Shift+Space global shortcut; real provider calls and... 36 分钟。

:speech_balloon: /so.nus/ 适用于 Node 的 STT（语音转文字），支持离线唤醒词检测

测试发现：With the recorded workarounds, the project's dependencies install, the snowboy native addon and grpc load, ESLint passes, the snowboy detector fires the hotword on real audio, and sonus's hotword-to-recognition flow works against a mock speech recognizer. 17 分钟。

一个开源的跟读文档阅读器服务器，提供高质量 TTS 选项、同步高亮显示，并支持将 EPUB、PDF、DOCX、TXT 和 MD 导出为有声书。

测试发现：Node.js v22.12.0 with pnpm 10.33.4 installed, all 810 unit tests pass (0 failed), TypeScript type checking clean for both main app and compute-worker, production build compiles successfully, linting and boundary checks pass, and the development server starts and serves all pages (root, signin, signup) with HTTP 200. 8 分钟。

用于高度表现力语音和逼真零样本语音克隆的 ComfyUI 节点

测试发现：ComfyUI-VoxCPM v2.4.0 loads in ComfyUI 0.42.0 on CPU; all 6 custom nodes registered under audio/tts categories; test suite runs 61/62 passing; ComfyUI server starts and responds on --cpu flag; frontend JS extension served; model scanning, settings, validation, and configuration types all verified standalone. 9 分钟。

语音 AI 运行时。本地优先的转录、说话人分离、TTS 和语音克隆，提供兼容 OpenAI 的 API。

测试发现：Izwi CLI 0.1.0-beta-18 and HTTP API server compile and run on linux-x86_64 with CPU backend; all 3169+ unit tests pass across 7 workspace crates; server serves OpenAI-compatible /v1/ endpoints on port 8080 with proper health, metrics, model listing, OpenAPI docs, and web UI; desktop Tauri shell cannot build due to... 30 分钟。

隐私优先的 AI 会议纪要工具，适用于 Windows 和 macOS。免费、开源（MIT）。设备端 Whisper/Parakeet 转录和说话人分离，无会议机器人。开源替代品，替代 Granola、Otter 和 Fireflies。

测试发现：Meetily Tauri app builds from source, all Rust (188) and frontend (8) tests pass, and the desktop app launches successfully on Xvfb, initializing all subsystems including Tauri GUI, notification system, database, tray icon, and model managers. 61 分钟。

ElevenLabs API 的官方 JavaScript (Node) 库。

测试发现：The repository installs with npm, builds with tsc, and its full Jest suite (1094 tests, 179 suites) plus a 22-check SDK probe pass under Node 20 against a local schema-correct mock of the ElevenLabs API, with generated mp3 audio verified by ffmpeg decode and WebSocket realtime/speech-engine paths verified live. 19 分钟。

开源的本地优先视频编辑器，创作者和 AI agents 可在其中编辑同一个真实时间线。

测试发现：npm install, TypeScript typecheck, and Vite build all pass; the dev server starts and serves the application at localhost:5173; the headless timeline CLI command engine parses commands and returns structured JSON errors. 9 分钟。

一个用于通过 OpenAI Chat API 与 GPT 模型交互的 Alfred 5 Workflow，还支持图像生成/编辑/理解、语音转文本转换以及文本转语音合成

测试发现：Ruby 3.4.7 installed via Conda; all 21 source files pass syntax check; 11 of 12 test files pass (the lone failure is a macOS-only afplay assertion); the WEBrick HTTP server boots and serves pages correctly; the HTML page builder produces complete output. 6 分钟。

一种能够一次性生成超逼真对话的 TTS 模型。

测试发现：Package nari-tts installs and all modules import, config loads from HF, CLI help parses, and audio utility functions work correctly, but the 1.6B-parameter model cannot load on this container's 755 MB of RAM (needs at least 3.2 GB). 10 分钟。

Speech Note Linux 应用。一款支持离线语音转文字、文字转语音和机器翻译的笔记、阅读和翻译工具。

测试发现：CMake configured successfully in build_final2 with Qt6 6.4.2, Catch2, fmt, ffmpeg; full build fails at dsnote_lib because 18+ external C/C++ API symbols from webrtc_vad, rnnoise, espeak-ng, piper, april-asr, sam, rdrview, cld2, google_pinyinim, xdo, qhotkey, libarchive, taglib, ssplitcpp, maddy, html2md, cpp-pinyin... 48 分钟。

自托管 AI 网关，用于私有 RAG、自然语言数据访问和工具调用 agent。

测试发现：无法验证；log 显示了它停在哪里。 28 分钟。

让 Azure 自然 TTS 语音可供任何 SAPI 5 兼容的应用程序使用。

测试发现：x64 release binaries (TtsApplication.exe and NaturalVoiceSAPIAdapter.dll) extracted from GitHub release v0.2.9 and run under a manually-extracted wine-9.0 environment with zlib1.dll provided from MSYS2 mingw64, the GUI app launches and stays running, and the COM DLL registers successfully via regsvr32. 20 分钟。

为 AI agents 赋予视觉、听觉和可验证的结果。Watch Skill 将视频、音频和屏幕活动转化为可搜索、带时间戳的证据，并用确定性契约而非模型意见来证明工作成果。DeepWatch 是基于 DeepSeek Harness 构建的 agent 工作区。Python +...

测试发现：watch-skill 1.4.3 installed with all extras; 2215 offline tests pass; ruff clean; CLI reports all dependencies operational; lint checks pass. 88 分钟。