How can I obtain the correct pitch from the partiels software.
From my experience any tymber I obtain a data file from a mubu the reply of those notes does not mutch the sound you hear from professional singer.
Thanks, for any help
Hello,
You can try to use the Feature Scoring, the FCN, or the Crepe analysis plug-ins. FCN and Crepe will be easier to use because there are very few parameters (they are based on neural network models). With the Crepe plugin, if you use a larger model, the results will be better, but the computation will be slower.
Best,
Pierre
Dear Guillet, you touched very important issue on partials. I would like to ask you a short excerpt of my wave sound using the function or programs you suggested me.
For me will a sample to use it for my purposes. If you also send me the picture of the preferences you used, that will help me to check if it works for my purposes.
Here is the short song of my studies.
With kind regards,
Berhane
Dear Pierre,
Here is a sample wave-sound, thnaks
(Attachment Nisgid-sample-ircam.wav is missing)
I can’t download the audio file. @IRCAM-Forum-team are there issues with the audio file on the discussion? Here is a screenshot of a document with the 3 analyses and a template document. Replace the audio file with yours and it should be okay.
Pitch.ptldoc (175.6 KB)
Hi @berane9,
As a new user, you were unable to upload any files in your message. I’ve just increased your trust level so you can share your file here. Can you try uploading it again?
Thanks
Thanks Beller. Berhane
Hi Guillot, it was interesting to discover an optmal way the value of Hrtz. I found the value of pitch I wanted to trace.
- Now I would like to know how to get the marker size, the time measure, so that I can associate the pitch with the size at the same point of intersection.
- I would like to know if there is an easy way to insert it with Max patch. Thanks
dear Guillot, on the place I found the crepe plugin the marker object exists, but it doesn’t function at all. So how can I make work is my question, if there are other process to enforce that task, dear. Thanks
Hello,
What do you mean by “marker size”? So I don’t understand what you want to do.
Otherwise, you can use the results with Max by exporting the data to a text file and loading it into the coll object for example, or you can use OSC to send the data directly from Partiels to Max in real time.
I meant for the “marker or partition”, the device that detects the time measure of note, how far the note will stay in play time.
This device will mark the dimarcation of time mater: how far a note will stay in a given time.
Hope I got to the point, and sorry.
You can look at the Transient Detection analysis, but it won’t correspond perfectly with the notes. Otherwise, you can try to segment with lyrics, so you can use the Whisper analysis to generate the lyrics and the VAX plugin to synchronize the lyrics on the audio (there is a template with the VAX manual).
Hey! I’ve noticed similar issues with pitch accuracy in Partiels before. One thing to check is if your calibration settings are correct—sometimes the pitch standard can be off. Also, the data from mubu might not capture the full complexity of a professional singer’s voice, especially in terms of harmonics and timbre. You could try manually adjusting the pitch or using post-processing tools to refine it. Hope this helps!