Skip to main content
During an active recording, Salesgraph Assist displays a live transcript powered by AssemblyAI’s streaming transcription service.

​
How It Works

Audio captured by the Recall AI Desktop SDK is streamed to AssemblyAI v3 for real-time speech-to-text processing. The transcript is speaker-attributed — each segment is labeled with the speaker’s name when identified. The SDK emits two types of transcript events:
  • Finalized segments — Complete, speaker-attributed transcript entries. These are the final version of what was said and do not change.
  • Partial segments — In-progress text that is still being processed. Partial segments update in real time as the speaker continues talking and are replaced by the finalized version once processing is complete.

​
Display in the Recording Widget

The recording widget shows:
  • The last three finalized segments, each with a speaker label and the transcribed text.
  • A partial segment displayed in italics below, updating in real time as words are recognized.
This gives you a running view of the conversation without overwhelming the compact widget interface.

​
Transcript Mapping

Raw transcript data from the Recall SDK arrives as word-level entries with timestamps and participant IDs. Salesgraph maps this into a simplified format:
FieldDescription
SpeakerThe participant’s name
TextThe full sentence or phrase
Start timeTimestamp when the segment began
End timeTimestamp when the segment ended
This mapping handles both real-time transcript data (during recording) and downloaded transcripts (after recording completes).

​
Full Transcript After Recording

Once a recording is completed and processed, the full transcript is available in the Meeting Detail view under the Transcript tab. The post-recording transcript includes all segments with speaker labels and formatted timestamps (mm:ss).