It can take a day or so to train a CNN model, or several minutes to create a profile, assuming that you have enough suitable samples. Fortunately it is possible to transcribe without prior training for a particular instrument. To illustrate this I have a single note played on a double bass (A octave 2).
![](http://sound-analysis.com/wp-content/uploads/2024/03/image-17-1024x261.png)
![](http://sound-analysis.com/wp-content/uploads/2024/03/image-22.png)
To transcribe I set the options to No Profile and Constant Q Mapping.
![](http://sound-analysis.com/wp-content/uploads/2024/03/image-18.png)
Selecting Use FFT Profile Result only
![](http://sound-analysis.com/wp-content/uploads/2024/03/image-19-1024x415.png)
The result is
![](http://sound-analysis.com/wp-content/uploads/2024/03/image-20.png)
Selecting Use Spectrogram Analysis Result only I get
![](http://sound-analysis.com/wp-content/uploads/2024/03/image-21.png)
Leave a Reply