Spec proposal: OVOS-AUDIO-1 — Audio Output Service

Proposal for OVOS-AUDIO-1, the audio output service specification.

Problem

The audio output service — the pipeline's output-side counterpart that consumes ovos.utterance.speak and renders responses as audio — has no normative contract. Topics, queue semantics, lifecycle signals, and stop integration are all implementation-defined.

Proposal

Defines the full contract for the audio output service:

• Rendering pipeline — dialog-transformer chain → TTS synthesis → TTS-transformer chain → playback queue
• Sequential playback queue — TTS speech and sound effects in declared order; ovos.audio.queue for scheduled sounds, ovos.audio.play_sound for immediate playback
• Output lifecycle — ovos.audio.output.started / ovos.audio.output.ended with session identity from context.session.session_id
• Speaking-status query — ovos.audio.is_speaking (session-scoped via context)
• Stop integration — ovos.audio.stop and universal ovos.stop; MAY scope to session
• Listen trigger — ovos.mic.listen emitted after playback ends when listen: true is set on the speak message

PR

PR #38

Read the proposal and PR #38 first, focusing on the listed topics, queue semantics, lifecycle signals, session identity, and stop integration. Done means the OVOS-AUDIO-1 specification has a complete normative contract for the rendering pipeline, playback, status queries, stopping, and listen triggering.