Open drscotthawley opened 12 months ago
Follow-up: Maybe something up with the website model's settings. When I run from the CLI, about twice as many notes are detected for the same audio recording.
Hi. The defaults in the website and python library should match, but if they don't, you can try to adjust the detection sliders in the website and see if there's a setting where your vocals come through.
@drubinstein How do we access those additional sliders on the website? I'm not seeing those anywhere:
Perhaps they only appear after the MIDI is generated? (which appears to be crashing right now?)
Yea they'll only appear after the midi is generated.
I've tried singing various vocal melodies and running them into BasicPitch, via the demo on the website. All solo vocals with no background noise. Live from the mic, or prerecorded via Audacity. Uncompressed or using dynamic range compression...
So far, it misses almost all of the notes I sing.
Was the model designed to ignore vocals?
Are there other... I don't know,.. suggested settings or modes to try when operating the model?