Dealing with "word spill" in Transcripts

If you are seeing the first word of a chunk included in the previous chunk, leading to jumbled and confusing transcripts, this is likely due to inconsistent active speaker event timestamps provided by the meeting platform.

While Recall doesn't have control over these inconsistencies, we do offer a parameter to handle this.

To help alleviate this issue, you can set the enhanced_diarization parameter to true when calling the Get Bot Transcript endpoint:

curl --request GET \
  --url 'https://us-east-1.recall.ai/api/v1/bot/{id}/transcript/?enhanced_diarization=true' \
  --header 'Authorization: <api_key>'

This enables additional diarization logic that better handles inconsistent speaker event timestamps and can improve the transcript quality for certain languages.