Audio Recording Skill
August 31, 2026 ยท View on GitHub
This skill records audio to a file. It needs ovos-dinkum-listener.
About
The skill records audio to a file and disables wake words and speech-to-text while it is active. It needs ovos-dinkum-listener.
A similar skill, OpenVoiceOS/ovos-skill-dictation, saves text transcriptions instead of audio.
To avoid trapping a user in recording mode, you can configure a stop hotword. This special wake word only works during recording mode. When detected, it restores the listener to its default state. By default, no stop hotword is set.
A recording started by this skill times out after 4 minutes. Change this limit with the max_recording_seconds setting.
If a mycroft.stop bus message arrives (for example, "stop" on the CLI), the skill takes dinkum out of recording mode, but only if this skill started the recording.
Dinkum does not yet have a native, optional timeout that uses voice activity detection to stop a recording after a period of silence.
Examples
- "new recording"
- "start recording"
- "new recording named {file_name}"
Entity hints
The skill ships locale/<lang>/intents/name.entity, a list of example recording titles ("voice memo", "meeting", "interview", ...) for the {name} slot in start_recording.intent. These are hints, not a closed list: any title you say, including one not on the list, still fills the slot and is used as the recording's file name; listed titles simply match with more confidence. ovos-workshop (>=9.5.0a1) registers every shipped .entity file automatically when the skill's language resources are loaded, so nothing needs to be configured for this.