# Whisperstream

> On-device speech-to-text for Windows & Mac, open source, offline

- **Website**: https://whisperstream.io/?utm_source=uneed&utm_medium=directory&utm_campaign=uneed-listing&utm_content=listing-page
- **Uneed Page**: https://www.uneed.best/tool/whisperstream
- **Pricing**: Free
- **Category**: Development
- **Tags**: Desktop Apps, Open Source, Dictation
- **Submitted by**: @lanreal
- **Launch Date**: 2026-11-04
- **Social Links**: [Twitter](https://x.com/djacksondev_) | [GitHub](https://github.com/lanreal-io/whisperstream)

## About

Free, open-source dictation for Windows and Mac

Whisperstream runs on Windows and Apple Silicon Macs (beta). Hold a hotkey, speak, and your words appear in the app you're using. Write emails, take notes, draft documents or dictate prompts without sending your microphone audio to a transcription service.

Speech recognition on your computer

Choose from NVIDIA Parakeet, Qwen3 ASR and Cohere Transcribe, with 39 dictation languages supported across the available models. Whisperstream recommends a model for your language, and you can choose a compatible alternative manually. Cohere Transcribe is an optional, larger download.

Local cleanup with Stream-1 Mini

Stream-1 Mini is our own fine-tuned local model for English Chat and Email dictation. It removes filler and applies spoken corrections before the text is inserted.

Its roughly 460 MB download and 4 GB memory recommendation make local cleanup practical on more machines. Setup checks its speed on your computer. You can also choose a customizable local model or configure an external enhancement provider.

Hear and answer your coding agents

Agent Speak connects to Claude Code and Codex. Read short summaries, hear agent responses through local speech, and type or dictate replies to your sessions. Choose spoken summaries or full responses and pause the feature whenever you want. The coding agents continue to use their own configured model services.

More ways to use your voice

- Push-to-talk and toggle shortcuts, plus live dictation preview.

- App-specific and website-specific cleanup profiles.

- Dictionary replacements and vocabulary Boost with CSV import. Boost uses AI enhancement.

- Searchable, encrypted local transcript history with audio playback.

- Audio-file import for transcribing recordings locally.

- Lockdown Mode for restricting enhancement to local processing and protecting history. Agent Speak pauses while Lockdown is active.

Free, with every app feature included

No subscription, trial timer, license key or required Whisperstream account. Dictation and local cleanup work offline after the models are downloaded. Optional cloud cleanup sends transcript text to the provider you configure, whose terms and fees apply.

Download Whisperstream · Source code · Documentation