Custom command output
Send the transcript to any script instead of pasting it. Pipe it into your own tooling.
Talk to your agents. Keep your keys.
Prompts are getting longer and more precise. Saying them is faster than typing them. Nota types into the terminal, the editor chat or the ticket, with the model and the key you pick.
Any model, your key
Pick from 30+ speech models, on your Mac or in the cloud. Keys sit in the Keychain and the audio goes straight to the provider. When a better model ships, switch that day.
You see the bill
Every model shows its cost per hour before you pick it. Home shows what you've spent this month, per provider. At 4 hours of speech a month, GPT Transcribe comes to about $1.10.
Modes per app and per site
A mode for the terminal, one for GitHub in the browser, one for Slack. Each has its own model, language and output. The right one switches on by itself.
Small things for builders
Respond output
Ask a question and get the answer in a panel, without pasting anything.
Custom endpoints
Any OpenAI-compatible transcription or chat endpoint. Cleanup can also run on a local Ollama.
Shortcuts actions
Start and stop dictation or switch modes from the Shortcuts app.
Say your next prompt.
The first 5,000 words are free. No account, no key needed to start.
5,000 words free
No account needed.
macOS 14.4 or later · Apple silicon and Intel