
Typd is a desktop dictation application that runs its speech model locally after a first download. Its source repository documents version 0.2.13 for Windows with supported NVIDIA hardware and Apple-silicon Macs. The app can paste dictated text and use spoken instructions to revise selected text.
Open application, restricted model
The desktop application code is GPL-3.0, but its downloadable model files have a separate license. The author allows free use inside Typd, including commercial work, but restricts redistribution, use in another product and extraction for use outside Typd without permission. The model is derived from Qwen3-ASR; those custom terms should not be confused with the original Qwen3-ASR base model's Apache license.
On Windows, the project says an NVIDIA GPU with at least 4 GB of memory is needed for the supported download path. It lists a 2.2 GB model download and additional NVIDIA libraries on first launch. Its maintainer notes that the early system can miss names or details; users should review important transcripts.
Local dictation and network behavior
The maintainer says dictation works offline after the model is downloaded. Packaged builds also check for updates and send version, OS and an anonymous install ID. Optional sharing of dictations is a separate feature that asks the user. Local inference alone does not mean every part of a packaged app is network-free.
Our repository profile links to the code and its terms. We have not installed or tested Typd.
EXPLORE THE SOURCE
Projects in this story
GitHub repositories
- SameRainbows/typd-app ↗View project details
Original reporting; primary sources reviewed November 14, 2026. No independent hands-on test claimed.



