Error: Failed to transcribe chunk 0: File transcription not supported. #39
Description
Describe the bug
Transcription from mic always fails with Error: Failed to transcribe chunk 0: File transcription not supported. Use microphone recording with transcribe_raw().
File transcription seems to work fine, however. See the logs below.
According to the verbose logs audio seems to be recorded, so I'm confused where the problem occurs exactly.
Environment
- OS: Arch Linux
- Whis version: 0.7.2
- Installation method: cargo install
Additional context
whis -v --print -f ~/test.wav
[verbose] Loading whisper model from: /home/user/.local/share/whis/models/ggml-small.bin
[verbose] Whisper model loaded successfully
[verbose] Unloading whisper engine from cache
Try a recording using FFmpeg and my laptop microphone.
whis -v --print
[verbose] Audio device: Default Audio Device
[verbose] Preloading whisper model in background...
[verbose] Loading whisper model from: /home/user/.local/share/whis/models/ggml-small.bin
[verbose] Audio device: 44100 Hz, 2 channel(s) -> resampling to 16000 Hz mono
[verbose] VAD enabled (threshold: 0.50)
[verbose] Whisper model loaded successfully
[verbose] Created final chunk 0 (9.9s)
[verbose] Recorded 157696 samples (9.9s at 16000 Hz mono)
Error: Failed to transcribe chunk 0: File transcription not supported. Use microphone recording with transcribe_raw().