No second AI subscription
You already pay for Claude, ChatGPT or another agent. OSTT runs Ghostwriter, its writing action, through that agent's CLI, so context-aware dictation costs nothing extra.
OSTT vs Superwhisper
Dictation that takes instructions, free. OSTT runs on your own AI agent subscription, reads the window when you allow it, and transcribes on your own machine.
The Short Answer
Superwhisper charges for context. OSTT gives it to you.
Superwhisper is the default dictation tool on fresh Omarchy 4.5 installs, set up to transcribe in the cloud. Its context-aware Super mode, Parakeet, cloud models and file transcription need a Pro license. OSTT does all of that without a license: local transcription with full GPU support, and context-aware dictation that runs on the AI agent subscription you already have.
# In Slack, say: "reply saying I can't make it"
# OSTT pastes the reply, written by your agent
# Local Parakeet on your GPU, free
ostt model local download parakeet-ultra
# Transcribe a meeting and summarize it
ostt transcribe meeting.mp3 -p summary -o notes.mdWhat OSTT Adds
What you get with OSTT, without a Pro license.
| Capability | OSTT | Superwhisper |
|---|---|---|
| Price | ✅ Free, no account | ⚠️ Free tier; Pro is $8.49 a month, $84.99 a year, or $249.99 once |
| Context-aware dictation | ✅ Free, per app | ⚠️ Pro: Super mode and context awareness need a license |
| Which AI writes | ✅ Your own Claude Code, Codex, OpenCode, Gemini CLI or other agent, on the subscription you already pay for, with its tools switched off | ⚠️ Superwhisper's hosted models, or your own API key on Pro |
| Screen context | ✅ Optional screenshot of the window, per app | ⚠️ App, window title, input text, selection and clipboard; no screenshot documented |
| Local Parakeet | ✅ Free, with full GPU support | ⚠️ Pro; local Whisper is free |
| Cloud transcription | ✅ Eight providers with your own key, billed by the provider | ⚠️ Pro |
| File transcription | ✅ Free: ostt transcribe | ⚠️ Pro |
| Shell pipelines | ✅ stdout, files, and processing through any bash command | ❌ Not documented |
| Open source | ✅ MIT | ❌ Closed source |
Checked against Superwhisper's plans and modes documentation and the Omarchy manual on October 24, 2026.
It sees what you see
Turn on the screenshot for an app, and the agent reads the conversation or document you are replying to, so names, terms and tone come out right.
Safe with what is on screen
The agent runs with all its tools switched off, so text in a screenshot or transcript cannot make it run commands or touch files.
Local by choice, not by upgrade
Parakeet and Whisper run on your machine with Metal, CUDA or Vulkan, free. No audio leaves your computer unless you pick a cloud model.
Open and scriptable
MIT licensed, with a CLI built for pipes: transcripts go to stdout, files, the clipboard or any shell command.
History you can retry
Every dictation is saved. Retry one with another model, or run a different action on it, without speaking again.
On Omarchy
Switch Omarchy's dictation to OSTT.
1. Add the plugin
omarchy plugin add https://github.com/kristoferlund/ostt-omarchy-dictate.git --enable2. Click the microphoneThe setup installs OSTT, helps you pick a model, and makes OSTT the dictation backend.
3. Keep your keysOmarchy's dictation keys stay the same. Hold Right Alt or F9 and speak.
4. Choose per appRight-click the microphone to give apps their own action and screen context.