Hello
I am completely new to PD and am coming to it for a very specific midi utility id like to create but have no idea where to start.
Im looking to create a tool that takes realtime mic input , analyzes all the spectrum with FFT and spits out midi . not pitch tracking, but the pitch and volume/pitch of each bin to be resynthesized outside of PD.

a basic example of what im talking about is what peter arblinger did with his "phonorealism" stuff, which he built a pd patch to do this exact thing. https://cdm.link/2009/10/the-speaking-piano-and-transforming-audio-to-midi/

i would love for any suggestions of patches that either already exist or resources to help me learn where to get started. thank you !

@benrossm [adc~] grabs a sound input from the microphone (if Pd preferences are set to use the microphone input).
[sigmund~] will take the output and send frequency and amplitude data....... as midi note and volume on a scale of 0-100 I think so....... multiply it [* 1.27] for a midi velocity control.
[makenote] will take that data and make a note-velocity-duration message...... and [midiout] will send it.

There are lots of other tools...... see the "top menu"... "help"... "list of objects" for objects that you can use to control changes you want to make before you send out the midi.
David.

P.S........ I should have mentioned that once you have "put" an object you can right-click it to get help with how to use it.
Also the order of connections is Important and you should read and understand this...... https://forum.pdpatchrepo.info/topic/13320/welcome-to-the-forum
Audio connections do not behave in the same way....... the audio thread is a continuous stream of data, but non audio connections.... control data.... need care.

hey thanks alot for that. i just got it sorta working. i didnt realize you could do it with sigmund. i was thinking that only did like monophonic pitch tracking and id have to use the fft object or something, but sigmund can track about 16 partials. which is good enough for now, i can always split incoming signal into more bands and add more sigmund objects if i need to. way easier than i expected.

i would rather setup a series of steep bandpasses which match exactly chromatic tones (or quarter tones),
FFT might save some CPU, but makes it very difficult to calculate the average power at a given nominal frequency.