VoiceStudio
Fully local voice cloning, dubbing and transcription in 646 languages
About VoiceStudio
VoiceStudio is an open source desktop application that does voice cloning, voice design, video dubbing, dictation, transcription and audiobook creation without sending anything to a server. It ships 16 text-to-speech engines, 11 speech recognition engines and a 646-language catalogue, and it runs on macOS, Windows and Linux, with a Docker image as well. The pitch is straightforward: the capabilities people currently rent from a cloud voice service, running on hardware you already own, with no account, no API key and no usage counter for the local workflow.
The practical differentiator over the hosted alternatives is not just privacy, it is the absence of a meter. Cloning a voice from a short clip, dubbing an hour of video into another language, or generating a full audiobook are all operations that get expensive per-minute in a subscription product, and here they cost whatever your machine costs to run. It also exposes an OpenAI-compatible local API, so code already calling a hosted speech endpoint can point at localhost instead of rewriting. Two caveats belong up front. The application is AGPL-3.0, which is a real obligation if you want to ship it inside a closed-source product or host a modified version, and the only route to different terms is an enquiry form with no published price and no online checkout. Separately, the speech models it downloads carry their own licences, which are not the application licence, so commercial use needs checking model by model. The project is pre-1.0 and moving fast, with the star count climbing steeply through 2026.
You install a packaged release for macOS, Windows or Linux, run the Docker image, or build from source with Node and Python. The desktop app opens on a launchpad that leads into the individual studios. Voice cloning takes a short reference clip, and the project says three seconds is usually enough, then mirrors that voice for new text. Voice design goes the other way: describe a voice by gender, age, accent, pitch and emotion and it builds one, with a gallery of ready-made designed voices to browse by style. Dubbing transcribes the source video, translates it, re-voices it, and keeps each speaker separate while lining the timing back up against the original. Audiobook and story workflows take a long script or an EPUB and cast different voices to different characters, chaptering the output. Transcription turns audio or video into editable, searchable text. Models are downloaded on demand from the in-app catalogue, and once they are on disk the whole thing works offline.
- •Local Voice Cloning - Mirror a voice from a reference clip of roughly three seconds, with nothing uploaded anywhere
- •Voice Design From a Description - Build a new voice by specifying gender, age, accent, pitch and emotion, or pick one from the voice gallery
- •Video Dubbing With Speaker Separation - Transcribe, translate and re-voice a video while preserving individual speakers and re-aligning timing
- •Audiobooks and Multi-Voice Stories - Turn a long script or an EPUB into a chaptered audiobook, casting separate voices per character
- •Dictation and Transcription - Real-time dictation plus audio and video transcription into editable, searchable text
- •OpenAI-Compatible Local API - Point existing speech code at a local endpoint instead of a hosted service, with 16 TTS and 11 ASR engines behind it
Developers and creators who are already paying per minute for cloud voice work and would rather spend the money on hardware, and anyone whose material cannot leave the machine for policy or privacy reasons - client audio, medical or legal recordings, unreleased content. The local API makes it a credible drop-in for a pipeline that currently calls a hosted TTS provider. Localisation work benefits most from the dubbing path, since speaker preservation and timing alignment are the parts that are tedious by hand. It is the wrong tool if you want zero setup and a hosted endpoint someone else keeps running, if your compliance story needs a vendor contract, or if you intend to embed voice generation in a closed-source product, because AGPL-3.0 and enquiry-only commercial terms make that a negotiation rather than a purchase.
Pricing
From $0/mo
Billed annually, no monthly price published
- Open Source$0/mo
- Pro$99/yr
From the vendor pricing page, 2026-10-08
