[Feature] Allow configuring / adding additional developer instructions for the new voice transcription feature
Resolved 💬 2 comments Opened Feb 26, 2026 by guidedways Closed Feb 27, 2026
💡 Likely answer: A maintainer (etraut-openai, contributor)
responded on this thread — see the highlighted reply below.
What variant of Codex are you using?
v0.106.0-alpha.8 CLI
What feature would you like to see?
Re: https://github.com/openai/codex/pull/3381 the new voice_transcription flag for voice transcription
The voice transcription API allows one to send a short prompt along with the audio for context. Please allow customizing (or adding to) the prompt so that we can include additional instructions. This would be useful in many cases, especially where one might speak in a foreign language but would want the transcription to always be done in English. It'll also be useful to those that work in a certain 'domain' and would want their transcriptions to infer / fill-in-the-blanks where the audio was deliberately terse. I've separately had great success with this using the API.
Additional information
_No response_
2 Comments
This feature is still under development. We're just playing around with it at this stage and haven't even committed to including it. We'll move it to "experimental" when it's ready for use and feedback.
@etraut-openai thanks - just to leave you with some feedback: it's been amazing so far! The only way I've been using the TUI now all day