# Kelados

> Kelados is a free, open-source (MIT) voice AI app that runs on your own device. It does voice dictation into any app, live meeting transcription with notes, audio/video transcription in 99 languages, and natural text to speech with 28 voices. Audio is never uploaded and there are no word, minute or character limits.

Key facts:
- Price: free. No subscription, no usage limits. Models run locally, so there is no per-use server cost.
- Privacy: speech recognition (Moonshine, Whisper) and speech generation (Kokoro-82M) run in the browser via WebGPU/WebAssembly, or in the desktop app. Audio never leaves the device.
- Platforms: web (any modern browser), desktop app for Windows and Linux (macOS coming soon), Chrome extension, Microsoft Word add-in, Node.js SDK/CLI, MCP server for Claude Code.
- Account: optional; only used to sync projects, presets, pronunciation rules and dictation snippets, and to issue API keys.
- Source code: https://github.com/LovekeshAnand/Kelados

## Products

- [Dictation](https://kelados.in/dictate/): real-time speech to clean, punctuated text. Removes filler words ("um", "uh"), understands spoken commands ("new line", "bullet point", "comma"), corrections ("4pm, no wait, 5pm" → "5pm") and "scratch that". Snippets and a personal dictionary expand as you speak.
- [Desktop app](https://kelados.in/download/): press Ctrl+Shift+Space in any application (Windows and Linux; Mac coming soon), talk, and Kelados types at the cursor. Works in Word, Slack, VS Code, browsers and terminals, including Claude Code. A free alternative to Wispr Flow.
- [Notetaker](https://kelados.in/notes/): live transcripts of Google Meet, Zoom and Teams calls without a bot joining. Your microphone is labelled "You", the shared meeting tab "Others". Produces highlights, action items, decisions and open questions; exports Markdown, TXT, SRT. Also a "my own notes" mode for dictated notes. A free alternative to Otter.ai.
- [Transcribe](https://kelados.in/transcribe/): transcribe audio and video files in 99 languages or translate them to English, with SRT/VTT subtitle export. Any file length.
- [Speech Studio](https://kelados.in/studio/): unlimited text to speech with 28 American and British English voices, streaming playback and WAV export. A free alternative to ElevenLabs for English narration (no voice cloning).
- [Projects](https://kelados.in/projects/): multi-voice audiobooks, podcasts and voiceovers exported as one WAV.
- [Voice library](https://kelados.in/voices/): previews of all 28 voices.
- [Integrations](https://kelados.in/integrations/): Chrome extension (Google Docs, Gmail, any text box; Alt+Shift+D), Word add-in, Claude Code.
- [Compare](https://kelados.in/compare/): Kelados vs Wispr Flow, Otter.ai and ElevenLabs.

## Developers

- [Developer docs](https://kelados.in/developers/): `npm install kelados`. Node SDK (`createKelados().speak(text, { voice })`), CLI (`npx kelados speak "text" -o out.wav`, `npx kelados say "text"`), a self-hosted HTTP server compatible with common TTS request formats (`npx kelados serve`), and an MCP server.
- Claude Code: `claude mcp add kelados -- npx -y kelados mcp` adds tools `speak`, `save_speech` and `list_voices`.
- npm: https://www.npmjs.com/package/kelados

## Optional

- [FAQ](https://kelados.in/#faq)
- Models: Kokoro-82M (Apache-2.0) by hexgrad; Whisper by OpenAI (MIT); Moonshine by Useful Sensors (MIT); runtime transformers.js by Hugging Face.