imitone community

Building a universal, human-friendly music interface. ...But first we have to fix the bugs!

You are not logged in.


Having trouble with imitone? Hit send on any Error Reports it creates, or use the in-app "send feedback" button.

#1 2016-09-11 12:41:49

Registered: 2016-09-11

On floor and ceiling. Errant Pitch tracking varience

Loved the floor ceiling blog post.

Really helped concretize what direction development was focused upon?

perhaps a more accurate comparison for Imatone's real world counterpart would not be a recorder?
But a Kazoo!
Ultimately accessible to anyone who can hum!

But to truly invest in a very high ceiling for imatone...
Perhaps Evan should be persuing input of jazz vocal masters? Taking advantage of the precedent/legacy an age of music has gained, instrumentalizing the voice!

Chiefly, I would imagine that advanced "scat" musicians would be the most proficient? Whose input and needs/concerns would ultimately describe what such a ceiling would require for Imatone to reach it's high ceiling goal?

Off the top of my head? An Imatone "dream beta team" would consist of:

Jon Hendrix.
The Manhatten  Transfer
And Bobby McFerrin
For a scat musician's perspective!

And lastly I would imagine Danny Elfman's input might be
Pretty illuminating! In consideration of his compositional habits. Where he always carries around a personal DAT recorder and hums any inspiration musical passage as it occurs to him! ( even if that means rushing to the restroom on a jet airliner to hum a few hundred lines of layered percussion notes... lest anyone consider him looney tunes! )
Considering how busy Danny Elfman is.
I would suggest screening Tim Burton's Planet of the Apes Behind the scene's DVD for a short documentary of Danny Elfman describing his "compositional humming strategies" in more detail.

Personally I find that the best in any field are accessible to any inquirey that shows an interest and respect towards their concentrations! (Admittedly, as much was loads easier before the internet )

Good Luck reaching that high ceiling.


#2 2016-09-11 13:27:43

Registered: 2016-09-11

Re: On floor and ceiling. Errant Pitch tracking varience

Seems from the illustration illustrating the problem of errant pitch varience woe...

That an editing stage during playback could easily allow one to "mark" each errant case.
In which case one could have a list of choices to best make a proper edit:

( the note is correct )
( merge the note to the same notes before or after unless both were the same )
( choose the note one was attempting )
( vibratto would have as a base, the note before or after instead of the errant note with a width parameter that would allow a bend of vibratto to reach either the errant note or past said note till an auditioned choice "becomes" pleasing.)
( The same as vibratto except the bend's width parameter effect is of a single stretch instead of a vibratto's frequency )
( one could visually mark all errant notes from the gui and choose to ignore or delete with a best guess on the deletions merge. Doing the same with vibratto could be auditioned? And might leverage the effect if then pleasing. )

Been reading feedback where end users were concerned with the fidelity of their microphones.
And I am wondering if dedicated/optional hardware could be researched for pitch capture?

I  have had some great results with similar capture tecnology using my own crystal piezo pickups to use with my Alesis D4 drum module. I made my Piezo pickups by simply rewiring Piezo buzzers easily purchased at the local radio shack with 1/4 line level jacks! ( piezo crystals then act in reverse as a velocity sensitive input device! )
Piezo is not just extremely attack velocity sensitive for capturing every nuance of percussive performance...
But My Roland GR1 Guitar synthesizer also uses piezo crystals in its pick ups (GK-1) That captures the guitar string's pitch, attack, sustain, bend, vibrato and as there is a cystal pickup under each string the entire device is multi timberal playing complex chords with a different character for each string depending on performance!

Seems that hardware research that complements an optional hardware purchase could go a long way in enticing musicians seeking a professional version of imatone with alot more power and control under the hood!


#3 2016-09-11 13:53:46

Registered: 2016-09-11

Re: On floor and ceiling. Errant Pitch tracking varience

Finally, an option to train Imatone to a personal vocabulary would probably be the best solution for modding an end result that mirrors an instruments character to that of ones performance peculiarities?

I would hope that there are now powerful voice recognition training dev libraries out now...
Considering how essential such sdk's will be for device acceptance? Particularly adoption of android and ios devices not to mention desktop voice search options from cortona siri and google search? And considering how complex a voice rec training technology would be without standard libraries?

But in addition to voice training...
I would like to be able to teach Imatone a musical performance library peculiar to whatever musical strategy I might have?

For instance Evan mentions quick attacks in his blog facilitating capture of percussive syllables like "duh"?

Why not explicitly train Imatone beginning a note with "duh" means quick attack?
A "duhmmph" would be duh +  mmph where mmmph could have been immatone trained to mute the note so the sum would equal the note muted with a quick attack?

Once again if we look towards the rich legacy jazz "Scat" musicians have left us it is easy to imagine building large complex libraries of vocal Instrumentation that would probably be impossoble for imatone unless particularly trained with a users own vocabulary strategy?


#4 2016-09-18 12:50:07

From: Olympic Peninsula, WA
Registered: 2014-05-23

Re: On floor and ceiling. Errant Pitch tracking varience

Hey, Claydough ---

Thanks for the enthusiastic feedback.  Point-for-point:

Kazoo:  Definitely a close relative in terms of being one of the most accessible instruments out there.  It's a shame it has a reputation for being obnoxious!  The key thing about it is that it changes the voice's timbre, eliminating one of the big factors that makes singing difficult.  The voice is very personal and putting it out there is sort of a vulnerable thing to do.

My latest work focuses on improving the way imitone interprets musicality and pitch, allowing it to more intelligently correct errors in the singer's tuning while matching their movements as closely as possible.  This is the other hard thing about singing, and if I can manage a nuanced solution I may succeed in creating the world's easiest instrument.

Professional vocalists / scat musicians:  That's certainly an interesting idea.  Historically, imitone's shortcomings have been easier for more experienced musicians to deal with, but I expect the final product to be vastly improved on both ends of the spectrum.  That said, the ceiling comes after the floor for what I'm doing and I'm inclined to hold off for now.

Editing:  Good thoughts here, but I think a note editor would be out of scope for imitone.  I wouldn't rule it out as a future project, but for the time being my goal is to improve imitone until it isn't making these kinds of mistakes.  And, actually, I've greatly reduced them in recent updates:  0.8.2 is much less "slippy" than its predecessors owing to a new "note interpreter", and the update I'm working on adds the long-awaited "hold mode" and "pitch guide" features which allow for far tidier recording (where desired).  I'm also likely slipping in a "quick-bend" function triggered by holding a key down.

Dedicated hardware:  I expect the final software to work well with most built-in laptop microphones.  An SM58 works wonders, though, and might be superior with any "high-sensitivity" options I decide to include.  I will say it would be awfully neat if imitone could be built into a decent microphone and sold as a hardware MIDI controller...

I've played with Piezos and microcontrollers a bit.  :)  I have a low-latency audio microcontroller called a Bela that I've been experimenting with.

Training technology / learning algorithms:  These are definitely A Thing, and I have an AI expert I chat with regularly...  But my approach to imitone doesn't really fall into that category.  That said, my next round of research may open a certain door related to what you're talking about.

I should note that there's a difficulty with training imitone to know what a sound "means":  How does it convey that to the receiving software?  Ultimately, imitone's output is generic MIDI messages, so while it would be possible to trigger different instruments (via different channels) with different types of sounds, imitone wouldn't really know what was on the receiving end and it would all involve some set-up.  That sort of thing is also a bit out-of-scope with my current plans, although it might be possible to do by using MIDI script plugins in combination with imitone studio's timbral articulations to choose where each note goes.

A more conventional way of using those timbral articulations would be to route them to a filter.  For instance, brightness usually maps to a low-pass on the instrument so that an open-mouth sound triggers a clear, sharp note while a closed-mouth sound makes the instrument sound dull.

Thanks again for sharing your thoughts.  A lot of this is out beyond my current horizons, but some of it isn't!  And it's fun to think about the far-flung stuff.  Feel free to share any other wacky thoughts you might have, and do check out the latest version!


Board footer

Powered by FluxBB