VoiceStudio
VoiceStudio is an open-source desktop application under AGPL-3.0 that runs voice cloning, voice design, video dubbing, dictation and batch audiobook jobs on your own hardware. It is a workflow shell over 16 speech-generation engines and 10 transcription engines, with a local API and an MCP endpoint for agents. Its default engine, OmniVoice, has CC-BY-NC weights, so commercial output needs a different engine.
What does VoiceStudio do?
VoiceStudio describes itself as an "open source voice cloning and workflow engine" whose tagline is "Build local." Its README organises the features into three groups: create (clone a voice from a clean reference recording, design a voice, dictate with a floating widget), produce (dub videos with timed speech, make stories, audiobooks and batch jobs) and connect (a local API and an MCP endpoint for agents, with optional remote workers).
The application itself contains no model. Its feature catalog lists 16 speech-generation engines, with the default "VoiceStudio (default, powered by k2-fsa/OmniVoice)", alongside CosyVoice 3, KittenTTS, MLX-Audio, VoxCPM2, GPT-SoVITS, sherpa-onnx, IndexTTS 2.5 and others, and 10 transcription engines with WhisperX as the default, plus Faster-Whisper, Parakeet TDT, Moonshine, FunASR and an OpenAI-compatible option. The catalog notes that "engine availability depends on installed models, hardware, and configured providers." The catalog also lists vocal isolation, speaker diarization, a batch queue, an AI watermark and GPU auto-detection.
Key facts
- Licence: AGPL-3.0-only on the Tauri shell, React front end, FastAPI backend and build scripts, per LICENSE-NOTICE.md; the bundled omnivoice Python package is Apache-2.0 from its upstream authors.
- Default engine weights: the k2-fsa/OmniVoice model card says its code is Apache-2.0 and its pretrained weights are CC-BY-NC because of training data constraints; the card lists 646 languages and 0.6 billion parameters.
- 16 speech-generation engines and 10 transcription engines are listed in docs/feature-catalog.md; other engines' licences are not stated there.
- Local API and MCP endpoint for agents; agent skills install with npx skills add debpalash/VoiceStudio.
- Downloads from GitHub Releases with guides for macOS, Windows, Linux and Docker; from source with bun install and bun run dev.
- The README says local workflows run on your hardware, remote services are optional, and usage analytics requires consent.
- A commercial licence for embedding VoiceStudio in a closed-source product is described in LICENSE-NOTICE.md, with pricing "coming soon".
- Latest tagged release when read: v0.5.3 on 2026-09-16; 32,494 stars on 2026-09-17.
What does it replace, and where does it fall short?
VoiceStudio replaces ElevenLabs for someone who wants cloning, dubbing and voice design in one desktop workflow on their own machine, with the choice of engine left open. Where Chatterbox is one model, VoiceStudio is the workbench that runs Chatterbox-class models and the transcription that dubbing needs, and exposes it all to agents through MCP.
Where it falls short: the licence of the output depends on the engine you pick, and the default one, OmniVoice, forbids commercial use of its weights. The README says it plainly: "Models have their own licenses; review them before commercial use." Hardware needs vary by engine and are not stated in the README, and the desktop app is mid-move from Tauri to Electron. For a hosted API or a single model with a clear commercial licence, see the open-source alternatives to ElevenLabs page.
How does VoiceStudio run?
Download a release for macOS, Windows or Linux, or run the Docker image; each has its own guide in the repository. From source, bun install and bun run dev start the Electron app. The README says "Hardware needs vary by engine; see performance" and gives no minimums itself. Engines and models download on demand. There is no hosted VoiceStudio; the maintainers link Ko-fi and PayPal for support and sell sponsor placements in the README.
Who is VoiceStudio for?
A creator or a small team doing dubbing, audiobooks or voice design who wants every step local and wants to switch engines as better ones appear. Anyone who needs one model with an unambiguous commercial licence should pick Chatterbox or Kokoro FastAPI directly.
What limits does the README state?
From the README and licence notice: models have their own licences to review before commercial use; the next desktop release ships Electron with one final Tauri sunset update; hardware needs vary by engine. And its own rule for cloning: "Clone voices only with permission."
Questions people ask
Is VoiceStudio open source?
Yes. The application code, meaning the desktop shell, the React front end, the FastAPI backend and the build scripts, is AGPL-3.0-only, an OSI approved copyleft licence. The bundled OmniVoice Python package is Apache-2.0. The models each engine downloads carry their own licences, which the README tells you to review before commercial use.
Can VoiceStudio output be used commercially?
It depends on the engine. The default OmniVoice engine's weights are CC-BY-NC, so its output is non-commercial. Other engines in the catalog carry other licences, and the README says to review each before commercial use. A separate commercial licence exists for embedding VoiceStudio itself in a closed-source product.
Does VoiceStudio run offline?
Yes. The README states that local workflows run on your hardware and that remote services are optional. Models download once per engine. The README adds that usage analytics requires consent, and that remote workers are an optional feature rather than a requirement.
How does VoiceStudio compare with ElevenLabs?
VoiceStudio covers cloning, dubbing, voice design and dictation on your own machine with any of 16 engines, and exposes it to agents through MCP. ElevenLabs is a hosted service with its own voices and a commercial licence on every output. The open-source alternatives to ElevenLabs page places VoiceStudio beside the single-model options.
Can agents use VoiceStudio?
Yes. The README lists a local API and an MCP endpoint for agents, and an installable set of agent skills with npx skills add debpalash/VoiceStudio. Remote workers are optional for offloading generation to another machine.
Sources
- VoiceStudio README, LICENSE and LICENSE-NOTICE.md: github.com/debpalash/VoiceStudio, read 2026-09-17.
- Feature catalog: docs/feature-catalog.md, read 2026-09-17.
- OmniVoice model card, Apache-2.0 code and CC-BY-NC weights: huggingface.co/k2-fsa/OmniVoice, read 2026-09-17.
Compared with the others
On the open-source alternatives to ElevenLabs page, VoiceStudio is the pick for cloning, dubbing and design in one desktop app. AGPL-3.0 workflow engine over 16 speech engines and 10 transcription engines, with a local API and an MCP endpoint for agents, all on your own hardware.
Also on that page: Chatterbox for voice cloning you can ship, Kokoro FastAPI for a speech api on your own server, Piper for small devices and offline assistants.
More text to speech and voice cloning
A multilingual text-to-speech system with voice cloning from a 10 to 30 second clip and inline emotion tags, released with its weights under a research licence that requires a separate agreement for commercial use.
A family of text-to-speech models from Resemble AI under MIT, with zero-shot voice cloning from a reference clip and a multilingual model covering 23 languages.
A research text-to-speech model with zero-shot voice cloning whose code is MIT but whose pretrained weights are non-commercial, because of the dataset they were trained on.
A fast, local neural text-to-speech engine maintained by the Open Home Foundation, used by Home Assistant and the NVDA screen reader, under GPL-3.0.
A Docker image that serves the 82-million-parameter Kokoro model through an OpenAI-compatible speech endpoint, with streaming, voice mixing and nine languages, all under Apache-2.0.
Tencent Hunyuan's 1.5-billion-parameter speech foundation model under MIT for code and weights, covering zero-shot text to speech, speech editing, enhancement and separation through one natural-language instruction interface.
Added September 17, 2026. Every claim above comes from the project's README, LICENSE or model card, read on September 17, 2026, or from the GitHub API on the date shown in the panel. Found an error? Write to reveneau@licheo.com and it is fixed in the next weekly pass. Repository: github.com/debpalash/VoiceStudio.
Running one of these inside your own environment, with your own data and your own security rules, is the kind of work Reveneau does. Read how a forward deployed engagement works.