Transcripts
Working with Structured Transcript Data
The transcripts endpoint exposes machine-generated verbatim records of participant sessions in a structured JSON format. Each transcript object contains:
Speaker turns — identifies participant vs system/moderator.
Timestamps — millisecond-level markers for start and end of each segment.
Text content — normalized speech-to-text output.
Developers can:
Fetch the full transcript for a given responseId.
Use query parameters (e.g. start, end) to limit results to a time window.
Combine with video or audio metadata for playback sync.
Paginate long transcripts efficiently.
This endpoint is designed for downstream automation — powering search, summarization, highlight extraction, or integration into custom analysis pipelines — without requiring raw media processing.
Transcripts are available within each TestSessionRecording
, which can be queried for via a TestResultNode
.
Fetch a specific response transcript
Here we'll start our query at the the userTestResult
resolver and from there select the recordings we're interested in, along with their transcripts.
Fetch transcription for all responses to a specific test
Since we can fetch many responses for a single test, we're also able to fetch all transcripts at the same time.
(For info on pagination, see Pagination)
Transcript data example
The transcript is returned as a JSON sting which will require decoding in your script itself.
It includes the entire transcript and also breaks the transcript down into individual sentences and words with individual timestamps.
Here's an example transcript, trimmed for brevity: