#I’m building an iOS app using the ElevenLabs Swift SDK and trying to implement audio-text synchroni

1 messages · Page 1 of 1 (latest)

noble pilotBOT
#
AI-generated Answer for .tamba.

Short answer: there’s no documented toggle in the dashboard yet, and there’s no published API field you can set for Agents to emit alignment over the Swift/WebRTC path. The SDK callback exists, but the server-side alignment payload isn’t documented as available to Agents over WebRTC.

What’s documented today

  • The Swift SDK exposes onAudioAlignment in ConversationConfig for character-level timings. This is shown in the official SDK README. However, it relies on the server sending alignment in audio events.
  • The Sep 29, 2025 changelog says “Agents now have a flag to enable alignment data in audio events,” but it does not say where to toggle it or how to set it via API. There’s no accompanying UI/API doc page that names the field.
  • In the Agents “Client events” doc, the audio event is explicitly not sent over WebRTC (audio goes through LiveKit). Since Swift SDK uses WebRTC by default, any alignment that would piggyback on audio events isn’t documented as being delivered to Swift clients.
  • Alignment is fully documented for TTS, not Agents:
    • TTS WebSocket has a sync_alignment query parameter that includes timing data with every audio chunk.
    • TTS HTTP “convert-with-timestamps” returns alignment with the generated audio.
    • Forced Alignment API aligns existing audio to text if you want post-hoc timings.

What this means for your app

  • There is no published dashboard toggle or API property (in the public docs) to enable alignment for Agents. The Swift onAudioAlignment callback won’t fire if the server doesn’t send alignment, and the docs do not yet show that happening over the Agents/WebRTC channel.

Practical workarounds (documented)

  • If you need progressive text reveal now, use one of the TTS endpoints that do return alignment:
    • Realtime TTS WebSocket with sync_alignment=true for inline timings during playback.
    • TTS convert-with-timestamps for one-shot generations.
    • Forced Alignment if you already have audio and a transcript.

Recommendation

  • Since the changelog mentions the feature but the flag isn’t documented in either the UI guides or the Agents API reference, open a support ticket to confirm availability and whether it can be enabled on your agent. Provide your agent ID and that you need “alignment data in audio events” for Swift SDK.

If ElevenLabs later publishes the exact field or UI location, it will appear in the Agents configuration docs or API reference; until then, the only documented path to alignment is via the TTS APIs listed above.

robust atlas
#

how do i open a support ticket?