If you compare the

secondsFromStart

value in the

message.artifact.messages

object with the actual value of when the message was uttered in the audio, it appears the transcription / message is 1 second ahead of the actual audio. In the image below, the cursor / line in the audio playback is at 9s but the message from the user has

"secondsFromStart": 9.559999

. I assume the

secondsFromStart

would correspond to the time from the start of the call at which point the user started speaking the utterance. Perhaps this

messages

object isn't a direct transcription of the audio but just the array of messages in / out of the agent and those timestamps may not actually correspond to the time of utterance in the audio. Call ID: c72f38f6-64c9-44ca-80a9-bd9efed1baee https://cdn.discordapp.com/attachments/1344678836695990314/1344678836859572356/Screenshot_2025-02-27_at_11.20.07_AM.png?ex=67c1c941&is=67c077c1&hm=36f74590c2a8b36b804bce2ef88c8fa2d872d5adf48fd4b325aea8a5bc1d5a29&

Actually I noticed that too, and thought it was a bug on my side.

@maxmetcalfe @VahidKowsari lLet me verify it and get back to you.