# Speech SDK > On-device text-to-speech with voice cloning and streaming speech-to-text for the browser. No server, no API keys. ## Docs - [Introduction](https://kucukkanat.github.io/speech): On-device text-to-speech with voice cloning and streaming speech-to-text for the browser. Models run on WebGPU in Web Workers — no server, no API keys, nothing leaves the device. - [Getting started](https://kucukkanat.github.io/speech/getting-started): Install the packages, add the Vite plugin, and speak your first sentence. - [Playground](https://kucukkanat.github.io/speech/playground): Every feature of the SDKs, live in your browser. Pick a model, clone a voice, speak with karaoke highlighting, and dictate. ## Packages - [Text-to-speech](https://kucukkanat.github.io/speech/packages/tts): On-device, voice-cloning text-to-speech for the browser. Runs Chatterbox on WebGPU in a Web Worker: no server, no API key, nothing leaves the device. - [Speech-to-text](https://kucukkanat.github.io/speech/packages/stt): On-device, streaming speech-to-text for the browser. Runs Moonshine or Voxtral Realtime on WebGPU (or WASM) in a Web Worker: no server, no API key, nothing leaves the device. - [Voice library](https://kucukkanat.github.io/speech/packages/voices): A persistent voice library for @kucukkanat/tts: voices live in IndexedDB, come with two bundled demo voices, and remember their encodings so a saved voice speaks instantly after a reload. - [React hooks](https://kucukkanat.github.io/speech/packages/react): React hooks for the @kucukkanat speech SDKs. Each hook takes the engine you created — no provider, no context, nothing hidden — and re-renders only when something visible changes. - [Audio utilities](https://kucukkanat.github.io/speech/packages/audio): Browser audio building blocks behind the speech SDKs, useful on their own: microphone frames, gapless playback, recording, and WAV decoding and resampling. - [Core](https://kucukkanat.github.io/speech/packages/core): Shared foundations of the @kucukkanat speech SDKs. As an app developer you'll mostly use three things from here: the bundler plugins (Vite, esbuild, Bun), typed errors, and capability detection. The rest (typed worker RPC, engine lifecycle) is for building engines of your own. ## Guides - [Voice cloning](https://kucukkanat.github.io/speech/guides/voice-cloning): Clone a voice from a few seconds of audio — recorded, uploaded or raw PCM — and speak any text with it, entirely in the browser. - [Bundlers](https://kucukkanat.github.io/speech/guides/bundlers): Set up Vite, esbuild, Bun, webpack or Next.js so the engines' Web Workers and demo voices ship with your app. - [Handling errors](https://kucukkanat.github.io/speech/guides/errors): Every SDK failure is a SpeechError with a stable code and a message you can show to users. - [Hosting and deployment](https://kucukkanat.github.io/speech/guides/hosting): What your pages need to serve the SDKs, how to self-host models for strict CSP or offline use, and how bundlers pick up the workers. ## Agent resources - [llms-full.txt](https://kucukkanat.github.io/speech/llms-full.txt): The full Markdown of every page in one file. - [Page Markdown](https://kucukkanat.github.io/speech/index.md): Append `.md` to any page URL to fetch that page as raw Markdown. - [JSON API](https://kucukkanat.github.io/speech/api/docs/pages.json): Page index of the JSON docs API; each entry links the page's JSON and Markdown forms. Described by the OpenAPI document at https://kucukkanat.github.io/speech/openapi.json. - [API catalog](https://kucukkanat.github.io/speech/.well-known/api-catalog): RFC 9727 linkset of the APIs documented here. - [AI catalog](https://kucukkanat.github.io/speech/.well-known/ai-catalog.json): ARD manifest of the agent-facing resources on this site (MCP server, skills, APIs). - [agent-readability.json](https://kucukkanat.github.io/speech/agent-readability.json): Manifest of every agent-facing artifact on this site. - [Sitemap](https://kucukkanat.github.io/speech/sitemap.xml): Every indexable page URL with its last-modified date.