Dictate and edit a message
Insert spoken text into an agent message, correct it, and send it when it is ready.
Last updated 17 September 2026
Insert spoken text into an agent message, correct it, and send it when it is ready.
Before you begin
Prepare Voice in Settings and open the intended agent chat or New Agent composer. Confirm the project and conversation before dictating, because the recognized text goes to that target.
The demo’s Keyboard shortcuts page distinguishes hold-to-dictate from continuous Hands-off dictation.
Steps
Open Settings → Keyboard shortcuts and search for dictation. Check the bindings for Dictate with voice and Hands-off dictation.
For one message, hold ⌘L while speaking and release it when finished. Wait for transcription to insert the text into the open composer.
Read the transcription. Correct names, file paths, punctuation, and any instruction that speech recognition misunderstood before sending.
Use the normal Send control after checking the draft. Avoid ending your spoken draft with “send” or “send it” if you want to review it first: Choro recognizes a trailing send command and can submit the message.
For continuous input, use ⌘⇧L to start Hands-off dictation and use the same shortcut to stop. Keep an eye on the selected conversation and the listening state.
After stopping, check the final message and the conversation’s response. Dictation changes how you enter text; the agent’s provider, model, access level, and project still govern the request.
What to expect
Your spoken text becomes an editable agent message. You can review and correct ordinary dictation before deliberately sending it.
Troubleshooting
The screenshot shows the real shortcut controls, not a fabricated transcription. If Choro asks you to open an agent chat, select a valid composer before retrying. If you release the shortcut while the speech model is still installing, wait for installation and dictate again. Stop continuous dictation before switching tasks or discussing unrelated material near the microphone.