debpalash/VoiceStudio

What it solves

VoiceStudio is a local-first audio production suite that allows users to clone voices, dub videos, and create long-form audio content without relying on cloud subscriptions, API keys, or external accounts. It provides a unified interface for multiple text-to-speech (TTS) and automatic speech recognition (ASR) engines, ensuring that data and processing stay on the user's own hardware.

How it works

The software acts as a management layer over 16 TTS and 11 ASR engines. Users can switch between these engines based on their hardware (CUDA, Apple Silicon MPS/MLX, ROCm, or CPU) and specific needs (e.g., zero-shot cloning or low-memory footprint). It supports various workflows including:

• Voice Cloning: Generating speech from a short reference audio clip.
• Voice Design: Creating voices based on age, accent, and style instructions.
• Video Dubbing: A full pipeline that transcribes, translates, and synthesizes audio for videos while preserving speakers.
• Audiobooks: Tools for multi-voice scripts and EPUB/PDF import.
• Dictation: A system-wide widget for live transcription.

Who it’s for

It is designed for creators, podcasters, and developers who require private, offline, or high-volume audio production and want to avoid the costs and privacy concerns associated with hosted voice services.

Highlights

• Extensive Engine Support: Integrates 16 TTS and 11 ASR engines covering 646 languages.
• Local-First Privacy: All voices, projects, and outputs are stored locally by default.
• Cross-Platform: Compatible with macOS (Apple Silicon), Windows, and Linux, with Docker support.
• Advanced Audio Tools: Includes vocal isolation (Demucs), speaker diarization (Pyannote/WhisperX), and AI watermarking (AudioSeal).
• Developer Friendly: Provides a local REST/SSE/WebSocket API, an OpenAI-compatible audio API, and an MCP Server.