Cloud text to speech with voice cloning, no GPU needed
VoiceLabs is cloud text to speech with voice cloning: the speech models run on our GPU servers, so you need no graphics card, no drivers and no model download. Type or paste text, pick a voice or clone your own from a short recording, and the audio comes back to your browser tab or to your code. The seven engines are open-source models that we host for you.
Cloud or local text to speech
Neural text to speech comes in two shapes, and each is the right choice for someone. The difference is where the model runs, and so who supplies the hardware.
Cloud, like VoiceLabs
- The model runs on the provider’s GPUs; you send text and get audio.
- Nothing to install, and no graphics card or driver setup of your own.
- Works from a phone, a laptop with no GPU, a script or an AI assistant.
- The provider keeps the models running and updated; you do not track the projects behind them.
- Your text is sent to the provider to be spoken.
Local, on your own machine
- The model runs on your hardware, which for a good open model usually means a GPU with enough memory.
- You install the runtime, download the weights and update them yourself.
- Text and audio can stay on your machine, and it can work without a connection.
- You absorb the maintenance of every model you pick, as the engines page shows with dates.
Choose local when your text must never leave your machine or you need to work offline. Choose cloud when you want the same voices on every device without keeping a GPU and a model running yourself.
What runs in the cloud
All seven text-to-speech engines are built on open-source models and hosted by us. You pick one on every generation, so you can match the engine to the language, the voice and the speed you need.
- Qwen3-TTS10 languages
- Qwen3-TTS CustomVoice10 languages
- LuxTTSEnglish only
- Chatterbox Multilingual23 languages
- Chatterbox TurboEnglish only
- TADA10 languages
- Kokoro8 languages
Together they cover 23 languages. The language index says which engines speak each one, and the engines page lists the upstream project behind every engine, with its last commit and what we patch.
Voice cloning without a local setup
Cloning runs on our GPUs like everything else, so there is no local setup for it either. You give VoiceLabs a short recording of clear speech and it creates a voice on your account in seconds. Through the HTTP API the recording is between 2 and 30 seconds long, sent with the text it says, by a key that carries the voice:clone scope.
A cloned voice lives on your account, not in one tool: the studio, the API, the MCP server and the browser extension all list the same voices. Of the engines, Chatterbox Multilingual is the one that speaks all 23 languages and clones from a reference clip. Clone only a voice you have the right to use: the responsible use policy says what that means.
Four ways to use it
In the browser
The VoiceLabs studio is a browser tab: write, choose an engine and a voice, clone one, and download what you make.
From code
A REST API with an OpenAI-compatible speech endpoint, and the @voicelabs/sdk TypeScript package. API access is part of Pro. See the developer API docs.
In Claude, ChatGPT and Cursor
A hosted MCP server lets an AI assistant speak text in your voices, with an OAuth sign-in and no API key. The MCP server page has the server URL and the setup for each app.
Reading the web aloud
The read-aloud extension for Edge and Chrome reads a page or a selection in a voice from your account.
What it costs
One plan: VoiceLabs Pro, US$8 a month or US$59 a year, with a 7-day trial that asks for a card up front. It includes all seven engines, voice cloning, API access and 2 hours of generated audio a month (fair use). It is a flat subscription: no per-character fee and no usage bill. The pricing page has the detail.
Questions about cloud text to speech
- What is cloud text to speech?
- Text to speech that runs on a provider's servers instead of on your own computer. You send the text, the provider's models turn it into audio, and the audio comes back to your browser or your code. With VoiceLabs the models run on our GPU servers, so there is nothing to install and no drivers to set up.
- Do I need a GPU to run text to speech with voice cloning?
- Not with VoiceLabs. The seven engines are open-source speech models, and the hosted service runs them on our GPUs, so your browser never needs one and there is no model to download. Running the same kind of model yourself is a different path: you supply the hardware, install the model and keep it working as its project changes.The engines and their upstream projects
- How do I clone a voice in the cloud?
- Give VoiceLabs a short recording of clear speech and it creates a voice on your account in seconds; you then pick that voice for any text. Through the HTTP API, the clone endpoint takes a recording between 2 and 30 seconds long plus the text it says, and needs a key with the voice:clone scope. Clone only a voice you have the right to use.Responsible use
- Which languages and engines does it cover?
- Seven engines and 23 languages. One engine, Chatterbox Multilingual, speaks all 23 languages and clones a voice from a short reference clip; the others each speak fewer, and English is the only language every engine speaks. 13 languages have more than one engine and a page of their own comparing them.Every language and its enginesThe engines
- How can I use it besides the web studio?
- Through the HTTP API, which includes an OpenAI-compatible speech endpoint, and the @voicelabs/sdk TypeScript package; through the VoiceLabs MCP server, which lets Claude, ChatGPT and Cursor speak in your voices; and through a browser extension for Edge and Chrome that reads a web page or a selection aloud. Your voices are the same in all of them.The MCP serverThe read-aloud extension
- What does cloud text to speech with VoiceLabs cost?
- VoiceLabs Pro is US$8 a month or US$59 a year, with a 7-day trial that asks for a card up front. It includes voice cloning, all seven engines, API access and 2 hours of generated audio a month (fair use). There is no per-character fee and no usage bill.Pricing
- What leaves my computer when the speech is made in the cloud?
- The text you want spoken and, when you clone a voice, the recording you give it are sent to VoiceLabs to be processed. That is the trade-off against running a model locally. If your text or recordings must never leave your machine, a local model is the right choice; if you would rather not keep a GPU and a model running, the hosted service is.Privacy policy
VoiceLabs Pro is US$8 a month or US$59 a year, with a 7-day trial. Voice cloning and the API are included.
VoiceLabs is built on the open-source VoiceBox project (MIT), and that attribution is permanent. The hosted service, with its GPU inference, accounts and billing, is a commercial product from Devino Solutions.