Practical overview
When it fits
Use Sayit when spoken input should become text in the active Windows application, especially when you want control over transcription mode and how the result is cleaned up.
Who it is for
- Windows users who want to dictate directly into editors, chat apps, browsers or other applications.
- Writers who want to revise dictated text through editable cleanup prompts, hotwords or per-app rules.
- Teams or operators who want to run the documented backend themselves.
Documented uses
- Dictate text into the application currently in use.
- Turn rough speech into corrected, formatted text or retain a faithful transcript.
- Use spoken instructions to translate, tighten or rewrite selected text when context-aware writing is enabled.
- Choose local, direct cloud-provider or self-hosted speech-processing arrangements.
Requirements & limits
The project documents a Windows installer, a public trial server for a quick start, and an AGPL-3.0 license. Cloud API use involves providers configured by the user.
- Local mode keeps speech recognition on the PC; with AI cleanup off, the README says nothing leaves the device.
- Cloud API mode connects the PC directly to the ASR and cleanup providers configured by the user, without a Sayit server in the path.
- The self-hosted backend uses FastAPI, WebSocket streaming and Qwen3-ASR; GPU recognition requires NVIDIA hardware, with 16 GB or more of VRAM recommended for the default server model.
- Context-aware writing is off by default and can use text around the cursor; the README says password fields are skipped.
- The repository identifies the project as open source under AGPL-3.0.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Start dictating in a Windows app
Install Sayit, choose a voice engine and use its shortcut to insert a spoken transcript into an active app.
Before you start
- A Windows PC.
- The Sayit Windows installer from the project’s releases page.
Steps
Download and open the Windows installer.
Open Sayit and choose a voice engine; the README says the default public server can be used for a quick trial.
In another app, press the configured shortcut, speak, then press it again to finish. Hold-to-talk is available through a separate shortcut.
Expected outcome
Sayit transcribes speech and inserts cleaned-up text at the cursor in the active app.
Guide limitations
- The README describes local mode or adding personal cloud-provider keys for regular use.
- The project documents Windows as the application platform.
Deploy the Sayit backend with Docker Compose
Prepare and start the documented self-hosted backend using Docker Compose.
Before you start
- Docker Compose.
- Provider and deployment settings for the backend.
- For GPU speech recognition, an NVIDIA GPU; the project recommends at least 16 GB of VRAM for its default server model.
Steps
Clone the SayIt repository and change to its server directory.
Copy config.example.yaml to config.yaml and .env.example to .env.
Add provider and deployment settings to .env and config.yaml.
Run docker compose up -d --build from the server directory.
Expected outcome
Docker Compose builds and starts the Sayit backend as described in the project README.
Guide limitations
- The README identifies FastAPI, WebSocket streaming and Qwen3-ASR as backend components, with an optional OpenAI-compatible cleanup model.
- The project recommends consulting its server guide for configuration, deployment, security and API details.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 13 October 2026
- Original website title
- crosswk/SayIt · Official homepage title · retained in its source language