anyone know about PID control theory?

Is it doing the pitch detection you were working on? Then it's quite spot on! I didn't talk about it because it seemed really complicated -I actually read some stuff and it was kind of approachable, but they skipped "the details" every five lines as in "then you park the plane"... I think your axe-in-the-dark method goes well here

I always try to find things, but it happens to me too. I was playing some Tigran Hamasyan yesterday, thinking how sporadically he records, found out he made a new album last year. And it reminded me of Chick Corea, how he recorded every single thing that came to his mind -or his hands. They're otherwise quite close, but it seems to me that Corea never got bored, even when everything he touched (and he touched everything) always sounded like Corea. He just seemed to have so much fun playing, anything else was a plus. And it's not like you get bored but you go and do the thing -more like nothing goes anywhere, you discard stuff at a very low level of the mind, you don't even get to say "hey, but I liked that!" -it's gone. Tigran seems to get bored, so he goes piano solo until he feels like having a band again. And I think it's ok, especially if it doesn't pay the rent. One day you find something, you get an idea from someone else's work, or you just feel like it, and you're back. For a while

 

ah man that's that guy who died last week. i watched his tiny desk concert, him and his old timer friend going at it. you could tell they were having a blast. the first thing i ever saw of him was a Prince community page post to honor his passing, said basically he was to Prince what Prince was to everybody else

me, my brother and a friend have been jamming recently, and after some time we've come up with material for about 3 or 4 songs. but my friend is the type who likes things to circle back and nail things down, but in a way that means basically 'rehearsing', which is a big time yawn. i already wrote the damn part, what does it need practicing for? either develop it further or try to come up with something new. so it becomes almost something that check-out feeling, not nearly so bad though because there is still interplay involved, me coaxing him to stray out into deeper waters, me putting in effort to keeping on the ball, nailing section changes on the fly. usually the problem of it just 'not happening' comes when sitting down to write, for me. more so than just solitary practice. the low level mind rejecting the task is actually very accurate description of the mechanism of attention problems, something i know a thing or two about

anyway i think what that old man must have known, what Chick Corea must have done, was how to maintain a steady habit of music in service of fun. something tells me he managed to not lose sight of music being something to do to look cool to yourself and the universe, in the same way dancing is. i've made alright music in the lonesome chamber, its fully doable, but i think eventually the momentum runs down if you do it long enough

reaktor on the other seems to be very well suited for shut in behavior. as Bill Hader said when asked 'won't it cool off?' in an SNL sketch concerning a fake chemical pizza:

'if anything, its getting hotter...'

yeah, the poly pitch thing is a real pain in the ass, serves up constant reminders to why people typically don't bother with it. the victory feeling got cut short after rediscovering how much getting it through midi out paper shreds it. even before that, the sweet spot is volatile and signal dependent. though it undoubtedly it does have a sweet spot. almost everything else out there including my previous attempts would pitch detect transients and add false sour sounding high notes to chords. if and when its finally finished it'll be the last of the hyper-autism magnet projects, like a curse lifted

 

was it online reading? don't suppose there's a shot in hell you remember the link? sometimes the greatest little tips can be found peppered in there

 

Pretty interesting. Just seems to me plugin's time delay's correction would be needed. For real time application. The generated content seems to me delayed in respect to signal being extracted from.

But still. Great!!!!

 

Yeah... that. It goes up and down, but it's definitely not a steady beam of fun

I saved them

The most approachable one is some CCRMA guy's thesis -not the thesis itself, but his description of previous methods. The one in chapter 3 would be similar to this. What he describes there was presented in this paper with less friendly manners. Then there's this one, based on the first guy's method, trying to make it more accurate on transients. It's all not very readable, but maybe you get something out of them... Also, they're not focused on polyphony (they were originally working on speech resynthesis), but I couldn't find much on that. Guess the Melodyne guy has all that stuff under his bed (?)

 

actually, its not that bad. the fft is so overlapped to hell, sound starts getting to the output after around 100 samples, even though the frame size is longer. i think you're hearing the bad envelope to mask the bad sound

the annoying oscillators everything routed to couldn't really turn off yet, so i took the residual that was removed from the fft and used it trigger an envelope, to stop the oscillators being an overwhelming wall of sound, even though it lagged behind the signal. this sample is representative of the delay between the input and detection, which isn't great but not terrible either

its still a bad sound because its just the detection going to oscillators, but at least they turn off

 

that first one was where i learned about looking for spectral peaks! looks like i disregarded that bit about looking for minima last time haha

that transient one is great. square magnitude before summing is a keeper. i've done that before but didn't think to apply it. those are exactly the little tricks. i found a paper on polyphony once, and they had a very involved system for checking stuff peaks, like an iterative process. yeah, thanks anyway guys

the melodyne guy haha. i wonder how many bodies that guy has stashed too. melodyne is interesting, for all the things it can do it makes me wonder what the hell he uses. i mean i guess it'd have to be fft right there's nothing else for it to be. i bet he figured out some way to combine frame sizes for the analysis part

what's that crazy russian lady's site for research papers, research gate. well, damn it, looks like im not collegiate enough for her illicit website

 

Indeed.

I would expect it to use multiple sizes, or maybe an fft and a wavelet transform... The tricky part is still choosing the peaks and finding trajectories (and in Melodyne's case, looking for harmonic structure).

I wouldn't ask

 

you never heard of that crazy russian lady that voids copyright on academic papers and hosts them for the world? i guess people mostly use her site anyway even if their institution their own access to journals. its this really odd mix of crookedness/legitimacy that kind of bewildering. i didn't know you needed proof of being with some academic institution to get access,

anyway, i suppose it doesn't really matter how he does it, if his pitch detection sucks. ( may or may not be true but the point is i get to have an opinion now)

 

You mean Sci-Hub. Yup, everyone uses it.

There's this meme about it. Fun part -after the capture blew they actually made the article free... (only that article of course).

Cool! You changed something? It sounds quite stable. The false relation makes it more obvious when some notes hang, heh

 

why was i thinking research gate. even looked at the wikipedia page without it clicking that i was barking up the wrong tree

i remember that picture doing the rounds. what an un-self aware response to a meta situation. it seems there's nothing but convention propping the research publishing industry up. i guess they own the rights to everything they're published, but i don't see why there isn't some mass public repository that researchers would transition to publishing to instead, like how wikipedia operates for the common good and maybe begs for money periodically to cover hosting fees. how it is now is like they're running a dairy with cows they don't have to provide care for, and who milk themselves.

that test was run with it controlling an external string sampler with a long release time. maybe not the best for an objective side by side comparison, but listener fatigue is a consideration also. actually it is helpful because if there is a spurious false detection it sticks around long enough to be perceptible, which is probably what you heard. actual stuck notes can only mean one thing: time to nuke the site from orbit and revert to a previous save

the midi out module reallllly doesn't play games, which is funny because everything happening right before it is like a trapeze act in a russian circus

would you be willing to test it out, as a sanity check, probably tomorrow or in the next couple of days? this part of the project, tuning a system with as many plates spinning at once, testing it with varied input that may or may not be a good candidate for pitch extraction, while falling subject to confirmation bias both positive and negative can make staying objective almost impossible.

 

Probably because of

For maths and physics there's arxiv.org. For the rest... not much. Some cases are particularly irritating, like IEEE or ISO selling the standards' pdfs...

Yup. I think they're not spurious, they just extend a bit into a chord where they clash.

Of course!

 

you have one of those cool guy profiles where i can't access it to send you a private message. how should i send file

 

Weird! Don't even remember doing that. Just sent you a message, I guess you'll be able to answer there.

 

always thought it was cooler to have it that way. "sorry un-referenced and unspecified person, no need of you pulling me up to see what you think of me.

paule had to whitelist me about maybe four years ago so i could message him, under practically identical circumstances lol