The Autron Agent takes dictation. Press the microphone in the composer, speak, and your words are transcribed into the message box — ready for you to read over and send. It is dictation only: the agent does not speak back, and nothing is sent automatically.
Start dictating
Click the microphone in the composer, or press Option+V.
Allow microphone access the first time your browser asks.
Speak. A waveform and timer replace the text row while you record.
Choose Stop and transcribe when you are done, or Discard recording to throw it away.
Where it works
Two places: the chat composer, and a routine's Agent Instructions field — useful when you are describing a longer piece of recurring work.
What to expect
It knows the vocabulary. Transcription is biased toward Amazon advertising terms, so ACoS, TACoS, SP and ASINs come back written correctly rather than phonetically.
Three minutes per recording. At the cap it transcribes what you said rather than discarding it.
Text appends. Dictating with something already typed adds to it.
Nothing sends itself. You always press send.
Your audio is not stored. It is transcribed and discarded.
Troubleshooting
"Could not access a microphone" — the browser blocked access, or no microphone is available. Check the site permissions and that another app is not holding the device.
"Transcription failed — check your connection and try again" — the recording did not reach us. Retrying usually works.
"Recording is too large to transcribe" — record in shorter passes.