Saydraft is in development. The current Windows preview is available as an unsigned portable ZIP. Models are installed separately during setup.
Saydraft for Windows
0.1.1 · Windows x64 · 157.3 MB
Unsigned development preview · portable ZIP. Extract the whole folder and run Saydraft.exe. Speech models are downloaded explicitly during setup.
Download for WindowsVerify SHA-256
479938ff19281bacbbb9e9da5860cf7e56b174292a68fd5cc4f76ed6fd29a95f157,322,503 bytes
Release notes, checksums and all assets ↗
Start the portable app
- Quit an older Saydraft instance from its system-tray menu.
- Download the ZIP and extract the whole folder to a location you can write to.
- Run Saydraft.exe on Windows 10 or 11 x64. The portable includes its Python runtime.
- Choose your microphone, review and install a speech model, then try the private setup practice.
Keep all extracted support files together. The preview has no installer; optional AI cleanup is available in the Models setup step and can be left off to use rules.
Set up once. Then go offline.
Review the recommended model, size and license, then start its download in setup. Progress and pause/resume are available. Later launches reuse your installed files. The setup walkthrough covers microphone selection, models, shortcuts and a first practice draft.
Arabic, Bengali, Chinese, Hindi, Indonesian, Japanese, Korean, Malay, Thai and Urdu use the optional shared All languages model. Download it once; there are no separate language packs. Portuguese, including Brazilian Portuguese, is also covered by the default model. Review language and model choices before choosing a package.
Speech model downloads
These transfer sizes come from the pinned app catalog at build time. MB means 1,000,000 bytes. Verified cache entries can reduce what you download; disk use also includes installed files, temporary transfers and optional retained audio.
| Model or package | Transfer | License |
|---|---|---|
| Default: Parakeet v3 + language detector | 773.4 MB | CC-BY-4.0 + MIT |
| English max: Parakeet v2 | 661.3 MB | CC-BY-4.0 |
| Whisper base | 147.9 MB | MIT |
| Whisper small | 486.2 MB | MIT |
| Whisper medium | 1530.6 MB | MIT |
| All languages: Whisper turbo | 1621.7 MB | MIT |
| Optional Silero speech detection | 2.3 MB | MIT |
| Optional GTCRN quiet-speech enhancement | 0.54 MB | MIT |
The default package includes its language detector. Optional AI cleanup, live preview and editing features have separate model/runtime requirements. See local models for those choices and their limits. Selecting a language never starts a transfer.
Run from source
For development, you’ll need Windows 10 or 11, 64-bit Python 3.13, Git, a microphone and disk space for your selected models. In PowerShell:
git clone https://github.com/tanzir71/saydraft.git
cd saydraft
py -3.13 -m venv .venv
.venv\Scripts\python -m pip install -r requirements.txt
.venv\Scripts\python -m saydraftThe repository can move ahead of the published preview. Check the release’s source commit and SHA-256 before comparing builds. There is no winget or Scoop command for this portable preview.