Skip to content

fix: ensure audio array is mono and maintain float32 dtype in extract_audio_and_text function#9

Merged
karamouche merged 1 commit into
mainfrom
fix/inversed-channels
Jun 5, 2026
Merged

fix: ensure audio array is mono and maintain float32 dtype in extract_audio_and_text function#9
karamouche merged 1 commit into
mainfrom
fix/inversed-channels

Conversation

@karamouche

Copy link
Copy Markdown
Owner

Summary

  • Fixed a bug where stereo audio arrays (shape (samples, channels)) were passed downstream as-is instead of being collapsed to mono, causing shape mismatches in PESQ/SNR/NISQA scoring
  • Ensured the float32 dtype cast happens before the mono-collapse so the mean is computed in float

Type of change

  • Bug fix
  • New feature / preset
  • Refactor / cleanup
  • Docs / CI

@karamouche karamouche merged commit 2f59d97 into main Jun 5, 2026
6 checks passed
@karamouche karamouche deleted the fix/inversed-channels branch June 5, 2026 09:01
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant