-
Notifications
You must be signed in to change notification settings - Fork 215
Description
Looking for tts_aligned_transcripts feature enabled with Node.js SDK
Problem
When using LiveKit voice agents with Node.js, AI transcripts consistently arrive delayed relative to audio playback in staged/production environments. This makes it difficult to sync transcript UI with TTS audio.
Locally the timing is acceptable, but once deployed the transcript lag becomes noticeable.
LiveKit already provides tts_aligned_transcripts in the Python SDK:
https://docs.livekit.io/agents/multimodality/text/#tts-aligned-transcriptions
However, this capability does not appear to be available in the Node.js agents SDK.
Proposed solution
Expose tts_aligned_transcripts in the Node.js agents SDK, matching the Python feature.
Ideally this would provide:
- Word or sentence timestamps aligned to TTS audio
- Or transcript chunks emitted in playback-synchronized order
- Available through session history or agent events
Why this matters
This would enable:
- Accurate real-time transcript synced to audio
- UX improvements
Without alignment, transcripts feel delayed and less trustworthy to users.
Additional context
I’m using LiveKit agents with ElevenLabs TTS and LiveKit inference. The issue is most noticeable when running agents in staged/production environments.
I originally asked about this in the LiveKit Slack and was advised to raise this here as an enhancement request.