debpalash / VoiceStudio

debpalash / VoiceStudio

VoiceStudio (formerly OmniVoice-Studio) allows you to clone voices, dub videos, dictate, and produce long-form audio on your own hardware. It features 16 TTS engines, 11 ASR engines, a 646-language catalogue, and support for macOS, Windows, Linux, and Docker. There are no accounts, API keys, subscriptions, or usage meters for the local workflow. VoiceStudio(前身为 OmniVoice-Studio)允许您在自己的硬件上克隆声音、为视频配音、听写并制作长篇音频。它拥有 16 个 TTS 引擎、11 个 ASR 引擎、646 种语言目录,并支持 macOS、Windows、Linux 和 Docker。本地工作流无需账户、API 密钥、订阅或使用量计费。

At a glance

概览

VoiceStudio Workflows: Voice cloning and design, video dubbing, dictation, stories, audiobooks, batch generation. VoiceStudio 工作流: 声音克隆与设计、视频配音、听写、故事、有声读物、批量生成。

Language catalogue: 646 TTS languages; actual coverage and quality depend on the selected engine. 语言目录: 646 种 TTS 语言;实际覆盖范围和质量取决于所选引擎。

Engines: 16 TTS · 11 ASR · switch in Model Catalogue or with Ctrl/Cmd+E. 引擎: 16 个 TTS · 11 个 ASR · 可在模型目录中或通过 Ctrl/Cmd+E 切换。

Platforms: macOS 13.3+ on Apple Silicon · Windows 10/11 x64 · Linux x86_64 with glibc 2.39+. 平台: Apple Silicon 上的 macOS 13.3+ · Windows 10/11 x64 · 带有 glibc 2.39+ 的 Linux x86_64。

Compute: CUDA · Apple Silicon MPS/MLX · ROCm on Linux · CPU · optional remote workers. 计算: CUDA · Apple Silicon MPS/MLX · Linux 上的 ROCm · CPU · 可选远程工作节点。

Interfaces: Desktop app · local REST/SSE/WebSocket API · OpenAI-compatible audio API · MCP Server. 接口: 桌面应用 · 本地 REST/SSE/WebSocket API · 兼容 OpenAI 的音频 API · MCP 服务器。

Storage: Voices, projects, settings, and outputs stay on the machine by default. 存储: 声音、项目、设置和输出默认保存在本地机器上。

License: AGPL-3.0 application; downloaded models keep their upstream terms. 许可证: AGPL-3.0 应用程序;下载的模型保留其上游条款。


Install

安装

Download a package from the latest release, then follow the platform guide. 从最新版本下载安装包,然后按照平台指南进行操作。

PlatformPackageGuide
macOS 13.3+Apple Silicon DMGInstall on macOS
Windows 10/11 x64MSI; choose the current-user build when listed to install without admin accessInstall on Windows
LinuxAppImage, x86_64 with glibc 2.39+Install on Linux
DockerCUDA, ROCm, CPU, and worker-only GPU profilesRun with Docker

First launch creates a managed Python environment and downloads the default model. Later launches reuse both. 首次启动会创建一个托管的 Python 环境并下载默认模型。后续启动将复用这两者。

Note: On macOS, first launch needs a one-time right-click, then Open approval. Intel Macs cannot run the local Python backend; use a remote backend instead. 注意: 在 macOS 上,首次启动需要右键点击并选择“打开”进行一次性授权。Intel 芯片的 Mac 无法运行本地 Python 后端;请改用远程后端。


First voice

首次声音克隆

Launch VoiceStudio and open Voice Cloning. Add a clean voice sample. Three seconds works; 5 to 15 seconds usually gives a better prompt. Enter text, choose a language, then select Generate. 启动 VoiceStudio 并打开“声音克隆”(Voice Cloning)。添加一段清晰的语音样本。3 秒即可生效;5 到 15 秒通常能提供更好的提示效果。输入文本,选择语言,然后点击“生成”(Generate)。


Features

功能特性

  • Voice Cloning: Zero-shot synthesis from a short reference clip. 声音克隆: 通过短参考片段进行零样本合成。
  • Voice Design: Create a voice from age, accent, pitch, style, and delivery instructions. 声音设计: 根据年龄、口音、音高、风格和表达指令创建声音。
  • Video Dubbing: Transcribe, translate, preserve speakers, synthesize, and export video. 视频配音: 转录、翻译、保留说话人特征、合成并导出视频。
  • Stories and audiobooks: Multi-voice scripts · EPUB/PDF import · chapter rendering · .m4b export. 故事与有声读物: 多人脚本 · EPUB/PDF 导入 · 章节渲染 · .m4b 导出。
  • Dictation Widget: System-wide shortcut, live transcription, optional local-LLM cleanup. 听写小部件: 系统级快捷键、实时转录、可选的本地 LLM 清理。
  • Vocal Isolation: Demucs speech/background separation. 人声分离: 使用 Demucs 进行语音/背景分离。
  • Speaker Diarization: Pyannote and WhisperX speaker assignment. 说话人日志: 使用 Pyannote 和 WhisperX 进行说话人分配。
  • Batch Queue: Queue large sets of audio and video jobs with per-job progress. 批量队列: 对大量音频和视频任务进行排队,并显示每个任务的进度。
  • Model Catalogue: Install, remove, select, and route TTS, ASR, and LLM models. 模型目录: 安装、移除、选择和路由 TTS、ASR 和 LLM 模型。

Comparison

对比

VoiceStudio trades managed cloud compute for local control. VoiceStudio 用托管的云端计算换取了本地控制权。

FeatureVoiceStudioTypical hosted voice service
Best fitPrivate, offline, self-hosted, or high-volume workFast setup without local model management
Data pathLocal by default; remote features are opt-inAudio and text are processed by the provider
Cost modelFree software; you supply the hardwareSubscription, credits, or metered API use
SetupInstall the app and model weightsCreate an account and use the web app or API
PerformanceDepends on your engine and hardwareProvider manages compute and scaling
Offline useYes, after required models are installedUsually requires a network connection
CustomizationSource, engines, models, API, and routing are openLimited to provider options
MaintenanceYou manage updates, disk, and computeProvider manages infrastructure
特性VoiceStudio典型的托管语音服务
适用场景私密、离线、自托管或高负载工作无需本地模型管理的快速设置
数据路径默认本地;远程功能需手动开启音频和文本由服务商处理
成本模式免费软件;您提供硬件订阅、点数或按量计费 API
设置安装应用和模型权重创建账户并使用 Web 应用或 API
性能取决于您的引擎和硬件服务商管理计算和扩展
离线使用是,在安装所需模型后通常需要网络连接
自定义源码、引擎、模型、API 和路由均开放仅限于服务商提供的选项
维护您管理更新、磁盘和计算服务商管理基础设施