Buzz Transcription: Hugging Face Fix for Non-Whisper Models

A single focused pull request fixes Hugging Face model search returning empty results for compatible non-Whisper models. Pull request 1715 broadens discovery to supported architectures while filtering out unloadable models and stale search replies.

Duration: PT2M12S

Episode overview

This episode is a short developer briefing from Buzz Transcription.

It explains recent repository work in plain language.

  • Show: Buzz Transcription
  • Published: 2026-11-22T14:11:07Z
  • Audio duration: PT2M12S

Transcript excerpt

This excerpt keeps the crawler page concise. Listen to the episode or use the RSS feed for the full update.

Good morning, it's Sunday, November 22nd, 2026.

There's no broad trend in the codebase today — just one focused, high-impact fix to model discovery.

The issue is in the transcription dialog: if you picked Hugging Face as a source and typed a non-Whisper name, like O-rook Fourier, you got an empty list.

The reason, described in pull request 1715, is that the picker always applied a Whisper-only filter, even when the selected backend supports Qwen 3 A-S-R, M-M-S, Parakeet, VibeVoice A-S-R, and Fun A-S-R Nano.

That meant perfectly compatible models were effectively invisible in the user interface, and users exploring newer speech models outside the Whisper family hit a dead end.

The fix changes the search logic in three ways.

Nearby episodes from Buzz Transcription

  1. Speaker Identification and Plugin Fixes
  2. Speaker Identification Improvements
  3. Hallucination Flagging and Metadata Caching
  4. Weekly Recap - Preserving Builds and User Work
  5. Build Stability and Workflow Protection
  6. Weekly Recap - Long-Recording Reliability and Workflow Fixes
  7. Skip Already Transcribed Prefix Fix
  8. System Tray Controls and Skip Logic Fix