need more verbose error output

[stan41calpoly]

I downloaded the jfk.mp3 sample and it works pretty well. So trying a youtube video that I downloaded, I converted it to mp3 and try the same thing. No output, no error. Hmm, I convert the mp3 to 16000 samplerate which is the same as jfk.mp3. No output, no error.

Input #0, mp3, from 'jfk.mp3':
Metadata:
Software : Lavf59.27.100
Duration: 00:00:11.00, start: 0.069063, bitrate: 55 kb/s
Stream #0:0: Audio: mp3 (mp3float), 16000 Hz, mono, fltp, 55 kb/s, start 0.069062
Metadata:
encoder : LAME3.100

Input #0, mp3, from 'apt_pupil.mp3':
Metadata:
encoder : Lavf62.4.100
id3v2_priv.com.apple.streaming.transportStreamTimestamp: \x00\x00\x00\x00\x00\x00\x00\x00
Duration: 08:01:30.45, start: 0.069063, bitrate: 24 kb/s
Stream #0:0: Audio: mp3 (mp3float), 16000 Hz, mono, fltp, 24 kb/s, start 0.069062

I know the duration is much longer and if that's a problem, print it.

[danbev]

Could you show us the command you are using to start whisper.cpp?
And if possible could you attach the sample mp3 that you are using.

[stan41calpoly]

I use:
whisper-cli -m models\ggml-large-v3-turbo-q5_0.bin -f apt_pupil.mp3

I did notice that once it finishes init "whisper_init_state: compute buffer (decode) = 98.70 MB", it just exits as it if never even tried to process the mp3. The file is 83MB.

After cutting the audio to 5 minutes, it is working. My guess is this this is a size issue. Perhaps printing that and explaining what the limits are in the output will reduce confusion.

apt_pupil.mp3 is 08:01:30.45 long

[danbev]

Perhaps printing that and explaining what the limits are in the output will reduce confusion.

Yeah, that is a good idea. I'll take a look and see what can be done 👍