Improvements

• Medical domain-specific model (Enhanced Operating Point) for Dutch, Finnish, French, German, Spanish giving the highest accuracy for healthcare use cases. Refer to the documentation for more details.

  • Relative accuracy improvements (medical domain): Dutch (nl) - 70%, Finnish (fi) - 40%, French (fr) - 51%, German (de) - 36%, Spanish (es) - 63%

• New models (Enhanced Operating Point) for Dutch, French, Spanish with improved transcription accuracy

  • Relative accuracy improvements (general domain): Dutch (nl) - 16%, French (fr) - 3%, Spanish (es) - 4.5%

• Minor speed improvements for Enhanced Operating Point (all languages)

• Speaker Diarization – Improved speaker change detection accuracy for long audio streams (1+ hours)

Updates

• Speaker Diarization upper limit for max_speakers removed and default value set to unrestricted

• Speaker Identification now supports get_speakers API configuration to generate speaker identifiers

• Channel Diarization now supports labelling audio events with specific channels

• Orchestrator Version - orchestrator_version included in the JSON transcript now follows the format YYYY.DD.MM+CommitHash+ReleaseVersion