SayIt

Open-source Windows voice typing with local, cloud-API or self-hosted recognition and editable AI cleanup.

Visit website (opens in a new tab) Compare with other tools

SayIt is an AGPL-3.0 voice typing app for Windows built with Tauri, React and Rust. Press the shortcut (press-again by default, with a separate hold-to-talk shortcut), speak, and the cleaned-up text is inserted at the cursor in any app. It runs in three modes: local recognition with seven bundled GGUF models and GPU acceleration, cloud API mode using your own keys for Doubao, Qwen, OpenAI, Gemini, OpenRouter, MiMo, Groq Whisper or any OpenAI-compatible endpoint, or server mode against a SayIt backend you deploy with Docker. Every AI cleanup preset is editable and per-app rules switch presets automatically. Optional context awareness reads text near the cursor and, with text selected, turns speech into an editing instruction such as translate or rewrite; password fields are skipped. Hotwords bias recognition on Doubao, Qwen, server mode and local Qwen3-ASR.

Pros

  • Free and AGPL-3.0; runs fully local with bundled GGUF models or against your own API keys
  • Editable cleanup prompts with per-app rules, hotwords, overlay live captions and local searchable history
  • Self-hostable FastAPI backend with Docker Compose for teams

Cons

  • Windows only; there is no macOS or Linux build
  • Cloud recognition requires your own provider keys and is billed by the provider
  • Live captions while speaking only work with the streaming Doubao and Qwen providers
Platforms
  • Windows
Speech processing
Hybrid
Works offline
Yes
Languages
—

Pricing

Pricing model
Open source
Free tier
Yes

Features

  • Yes Works in any app
  • Yes Real-time transcription
  • Yes Auto punctuation
  • Yes AI cleanup
  • Yes Custom vocabulary
  • No Voice commands
  • Yes Push to talk
  • No Auto language detection
  • Yes Translation
  • No File transcription
  • No Meeting notes
  • Yes Multilingual interface
  • Yes History

Compare SayIt with

Last verified 2026-09-22