/voice
Records spoken input and transcribes it as a message in the chat.
/voice starts an audio recording via the microphone and then transcribes the speech into text, which is sent as a chat message. Format (--voice-format, default wav; webm/mp3 need ffmpeg) and language (--voice-language, ISO-639-1 code, default en) are configurable.
✅ WHEN TO USE IT?
Dictate a longer request faster than typing it
/voice to freely describe a detailed request for a new feature
Work hands-free, e.g. while reading code on a second screen
Use /voice to speak observations directly instead of switching between reading and typing
⛔ WHEN NOT TO?
Precise code identifiers, file paths, or special characters need to be exact
Speech recognition often mis-transcribes technical terms, CamelCase names, or paths.
Better: Type such details directly, or insert them via /paste instead of speaking them
No working microphone or supported audio environment is available
/voice needs microphone access and, for some formats, ffmpeg; without that, recording fails.
Better: Type the text normally, or paste it from the clipboard with /paste
SOURCES
- Aider Docs — In-chat Commands ↗ aider.chat