OSTT for Omarchy
Dictation that takes
instructions.
Talk to any app on Omarchy. Ask for a reply instead of just a transcript, and your agent reads the screen and writes it for you.
Free and open source MIT
Dictate
Speak into any app.
OSTT plugs into Omarchy's own dictation. Talk, and your words land where your cursor is, in any app.
Transcribed on your own machine. Parakeet or Whisper, no cloud.
Reproduced on 2.3.1. Exporting a file with a space in its name writes an empty file. Renaming it fixes the export, so the path is probably not quoted.
You, speaking
Reproduced on 2.3.1. Exporting a file with a space in its name writes an empty file. Renaming it fixes the export, so the path is probably not quoted.
Connect
Processing runs on the agent you already pay for.
Before your words are pasted, a processing action can clean them up, rewrite them or turn them into a reply. OSTT runs those actions on the coding agent you are already signed in to: no new account, no API key, and the agent's tools switched off.
- Claude Code
- Codex
- OpenCode
- Gemini CLI
- GitHub Copilot
- Cursor
- Grok
- Pi
- Hermes
- OpenClaw
Instruct
Say what you want written.
Ghostwriter is the processing action that takes instructions. Turn it on for an app, and OSTT sends your agent a screenshot along with your words. The agent reads the conversation and writes the reply, names and dates included.
# release
Can we still ship 2.4 on Thursday? QA found a crash in the export dialog.
The export fix is in review, should land tomorrow morning.
@kristofer are we good, and who's on the changelog?
Yes, Thursday works for 2.4 as long as Jonas's export fix lands tomorrow morning. I'll have the changelog ready tonight.
Message #releaseYou, speaking
Tell her yes if the fix lands in time, and that I'll do the changelog tonight.
Edit
Select and make changes.
Select a passage and say what to change. The agent sees the rest of the page, so the new text fits in.
release-notes.md
OSTT 2.4
- Agents: Gemini CLI and Copilot join the built-in agents.
- Popup: The level meter follows your microphone's gain.
- fixed exporting files that have spaces in the name, it wrote empty files beforeExport: Files with spaces in their names no longer export empty.
- Daemon: Starts with your session when installed as a service.
You, speaking
Rewrite this like the other points.
Choose
Choose what dictation does in each app.
Plain transcription in the terminal, Text cleanup in your notes, Ghostwriter in Slack. Right-click OSTT's microphone in Omarchy's top bar and pick a processing action for each app, and whether it sees the screen.
Dictation
Ghostty Plain transcription
Obsidian Text cleanup
Slack Ghostwriter
Typora Ghostwriter
Local and fast
Parakeet and Whisper on your CPU or GPU, kept loaded so dictation starts the moment you speak.
Never loses a word
If the agent fails or takes too long, your words are pasted as you said them.
Screenshots on request
Only in the apps you choose, sent only to your agent, deleted when the dictation is done.
Free and open source
No price, no sign-up, no subscription. MIT-licensed, configured in plain TOML you keep with your dotfiles.
Try it.
Install the plugin to make OSTT Omarchy's default dictation tool. OSTT is free and requires no account. The plugin is open source and MIT licensed.
maya opened this issue
Exported files come out empty for some projects. Anyone else seeing this?