Skip to content

Common Issues

Transcription crashes on Docker

If Whisper transcription hangs or exits unexpectedly in Docker, your CPU may lack AVX2 support. Switch to the legacy-cpu image tag. See Transcription → legacy-cpu.

Titles are inconsistent after transcription

Transcription models can vary wildly on capitalization, number formatting, and punctuation. For the most part that's just the nature of the beast, but there are a few things you can do:

  1. Use an English-only1 model for English audio, or a multilingual2 model for other audio.
  2. For non-English audio with Whisper, ensure you've selected a language (not Auto).
  3. Step up to a larger Whisper variant (tiny → small → turbo), or switch to Parakeet.
  4. Enable Bias Words and add book-specific names and terms.
  5. Run AI Cleanup as a post-processing step and let a machine do the work for you.

xHE-AAC books fail

These books are not currently supported in Achew. See Supported Formats.

First launch takes forever

Achew is downloading and installing Python dependencies and building the project. This process may take several minutes, but subsequent launches will be much faster.

  1. English-only models: Parakeet 0.6b v2, Whisper .en ↩

  2. Multilingual models: Parakeet 0.6b v3, Whisper non-.en ↩