At a Glance:
VoiceStudio is a local-first ElevenLabs alternative for voice cloning, design, dubbing, dictation, and long-form audio, with 16 TTS engines, 11 ASR engines, desktop apps, and local OpenAI-compatible and MCP interfaces.
Overview:
VoiceStudio is an open-source AI voice studio for creating, transforming, and transcribing speech on macOS, Windows, and Linux. Its workflows cover zero-shot voice cloning, designed voices, speaker-preserving video dubbing, system-wide dictation, multi-voice stories, audiobook rendering, and queued batch jobs. A model catalogue routes work across local TTS and ASR engines using CUDA, Apple Silicon, ROCm, or CPU, while optional remote workers can supply additional compute. Voices, projects, settings, and outputs remain on the machine by default, and network-backed features are explicit opt-ins.
Key Decision Points:
Platform requirements: Packaged releases support Apple Silicon on macOS 13.3+, Windows 10/11 x64, and Linux x86_64 with glibc 2.39+; Intel Macs require a remote backend.
Hardware responsibility: CPU operation is supported, but generation performance and memory needs vary by engine; the README recommends 16 GB RAM, 20 GB of SSD space, and 8 GB or more VRAM for accelerated multi-stage work.
Engine-specific coverage: The catalogue lists 646 TTS languages, but cloning, instruction control, platform support, quality, and licenses depend on the selected engine.
Local and remote boundaries: The desktop uses a loopback-only FastAPI backend by default, while remote workers and OpenAI-compatible ASR endpoints are optional and visibly move audio beyond the local machine.
Integration surface: Local REST, SSE, WebSocket, OpenAI-compatible audio endpoints, and an MCP server support speech generation and transcription outside the desktop interface.
Core Features:
Multi-engine voice generation: Select among 16 TTS engines for zero-shot cloning, preset voices, delivery instructions, and language-specific synthesis.
Video dubbing pipeline: Transcribe and translate video, preserve speaker assignments, synthesize replacement speech, and export the completed video.
Stories and audiobooks: Build multi-voice scripts, import EPUB or PDF documents, render chapters, and export
.m4baudiobooks.System-wide dictation: Trigger live transcription with a global shortcut and optionally clean the text with a local language model.
Batch and model management: Queue audio and video jobs, monitor per-job progress, and install, remove, select, or route TTS, ASR, and LLM models through the catalogue.
OpenAI-compatible and MCP access: Generate speech, transcribe audio, list voices, and expose synthesis or transcription tools to compatible clients through the local backend.
Use Cases:
Clone a speaker from a short clean reference clip and generate speech locally in an engine-supported language.
Dub a video by transcribing, translating, assigning speakers, synthesizing replacement audio, and exporting the result.
Produce multi-voice stories or audiobooks from scripts, EPUB files, or PDF documents.
Connect an OpenAI-compatible audio client or MCP client to local speech synthesis and transcription services.
Open-Source Alternative Value:
VoiceStudio is an AGPL-3.0 alternative to hosted voice services such as ElevenLabs, with downloadable desktop packages, Docker installation, and source-based browser or desktop workflows. Its local backend exposes the engine catalogue, voice profiles, TTS, ASR, dubbing, and long-form pipelines through documented APIs and MCP tools. Developers can modify registry-based TTS, ASR, and plugin interfaces, while deployments can keep the default workflow local or explicitly enroll remote workers when additional compute is required.
