Frequently Asked Questions:
Transcription Confidence Heatmap

Common Questions

Does a low confidence always mean the text is wrong?

No. It indicates lower model certainty, which may or may not correspond to an actual error.

Can confidence thresholds be adjusted?

Yes. Thresholds can be tuned to match different review strictness levels.

Is this used to automatically change subtitles?

No. The heatmap is a review aid and does not modify subtitle content.

Limitations

  • Confidence scores reflect model certainty, not semantic correctness
  • Highly technical jargon may still require domain knowledge review
  • Visual emphasis does not automatically correct transcription errors

Work with AI you can inspect and control.

  • Explainable AI decision making
  • Assists human judgment rather than replacing it
  • Consistent, reproducible results