Thanks — I’m responding about getting start/end timestamps for transcription segments when using the OpenAI realtime model. Start and end times are available, but only when you enable TTS-aligned transcription and handle timed transcript objects. To get timestamps: • Enable TTS-aligned transcripts for the agent session (set

use_tts_aligned_transcript=True

). • Implement a

transcription_node

that accepts

TimedString

(or iterables of

TimedString

) and reads

start_time

and

end_time

(seconds relative to the agent turn). Example:

async def transcription_node(self, text: AsyncIterable[str | TimedString], model_settings: ModelSettings) -> AsyncGenerator[str | TimedString, None]:
    async for chunk in text:
        if isinstance(chunk, TimedString):
            print(f"Text: '{chunk}', Start: {chunk.start_time}, End: {chunk.end_time}")
        yield chunk

Note:

TimedString

/ TTS-aligned transcription is experimental and requires the aligned-transcript feature to be enabled; without that, segments may not include timestamps. If you want, tell me how you’re creating your AgentSession and I’ll show where to set

use_tts_aligned_transcript

or review your transcription_node code. Sources: Text and transcriptions | LiveKit Documentation, with interactive elements