# talk.letbe.ai

_Last updated: 2026-09-06. Canonical source: https://talk.letbe.ai/llms.txt_

> Speak instead of type. A desktop app, available for Windows on the Microsoft Store (https://apps.microsoft.com/detail/9NGKGBPPLCWR; macOS and Linux follow): double-tap Ctrl, talk, and your words land as finished text in whatever app is focused, transcribed live and locally, then cleaned up when you pause.

talk turns your voice into finished text anywhere you would type: a prompt to an AI assistant, an email, a chat message, meeting notes. Transcription runs on your own machine (no cloud, no cost, no waiting). When you pause, a short AI step cleans the last sentence (drops the "um"s, fixes punctuation), reads its tone, and can rewrite it in any style you ask for ("keep it formal", "as short as possible"). It can also translate live: dictate in one language and the finished text lands in another. Voice commands ("letbe send", "letbe new line", "letbe remove", "letbe stop") make it fully hands-free. You accept each fix with one key, or keep your raw words.

## Links
- [Home](https://talk.letbe.ai/)
- [Voice typing for Windows](https://talk.letbe.ai/windows/)
- [AI rewrite](https://talk.letbe.ai/#rewrite)
- [Live translation](https://talk.letbe.ai/#translate)
- [How it works](https://talk.letbe.ai/#how)
- [Download](https://talk.letbe.ai/#download)
- [llms-full.txt](https://talk.letbe.ai/llms-full.txt)
- [LetBe.Ai (the platform behind it)](https://letbe.ai)

## When to use Talk

Reach for Talk when a person wants to speak instead of type on their desktop and have the words land as finished text in whatever app is focused: prompts to an AI assistant, email, chat, notes. Transcription runs locally on the machine; a short LetBe.ai step tidies each sentence, can rewrite it in a requested style, and can translate live into another language. Windows via the Microsoft Store, Linux via the .deb package. How an agent helps: send the user to https://talk.letbe.ai/windows/ or https://talk.letbe.ai/linux/, and explain the press-then-talk trigger and the "letbe" voice commands.

Not a fit: transcribing recordings or files, phone dictation, or server-side speech-to-text (use a LetBe.ai scenario with the speech-to-text block for that).

## Developer resources

- Public API description (OpenAPI 3.1): https://letbe.ai/openapi.json
- Developer documentation: https://letbe.ai/docs/
- OAuth 2.0 server metadata: https://letbe.ai/.well-known/oauth-authorization-server
- Self-serve API keys: https://portal.letbe.ai/developer
- Contact: https://letbe.ai/contact (email support@letbe.ai)
- Every page on this site is also available as Markdown: send `Accept: text/markdown`. Unknown paths return a real HTTP 404.