Imitone -- wow! Most embarassing post I ever started

[GreyLion]

I'd love to see an actual 1.x product at some juncture. You're still in early beta six years later on software that you cashed in on by selling it as just needing a little polishing. You have blown through every single deadline you've ever stated.

You have zero credibility at this point, and yet another aspirational speech like the one above won't cut it. Release something that does the thing that you promised it would do.

[interactopia]

I'd love to see an actual 1.x product at some juncture. You're still in early beta six years later on software that you cashed in on by selling it as just needing a little polishing. You have blown through every single deadline you've ever stated.

In 2015, I legitimately thought imitone was almost done. I was wrong.

We had a huge booth at NAMM and plans to finish within a year. I talked to some old-timers there... I learned some more about the history of pitch-to-MIDI, in the 80s, 90s, 00s... In the following years I even saw another pitch-to-MIDI product pulled from the showfloor. I learned about failure after failure after failure. It became abundantly clear to me that even with the best tech in the world, imitone 1.0 would be a total disappointment and disappear into the annals like all its predecessors.

I can't understand your assertion that spending six years working to develop imitone into a good product is somehow less honest than slapping a 1.0 on a bad product. The overwhelming majority of our customers and backers do not seem to share this perspective (although they're very eager for a more useful, portable product).

Anyway, I've learned one lovely lesson from all this: research and release dates don't mix.

Hi Evan. I have a decent voice to midi setup in Bidule that tracks pitch and amp envelope along with some other vocal properties and converts it all to midi note, pitch bend and two CCs of your choice. I do dig jam origin’s stuff. Unfortunately nothing, including you guys’ product is a simple (just like any plugin) setup for just singing or humming and getting a useable midi stream inside your daw. I’m sure it will get there.

Again, we have a pitch-to-midi VST. It uses default UI, though, so it isn't as handy as our app. We'll be making big UX improvements to it later this year.

Personally, I prefer virtual MIDI despite the inelegance of it — I find it much easier using the app/plugin as the DAW's MIDI controller than devising some ingenious MIDI routing scheme for a multitrack project... But you hardcore types can use the a2m plugin and do just that.

Waves’ new vocal synth could easily have a midi output added and take the monophonic part of this market over as it tracks better than anything else I’ve used (which is almost every pitch tracking plugin ever released).

Ah, there's a tricky detail here. It's a bit easier to make a vocal resynthesizer based on (relative) pitch tracking than to make a pitch-to-MIDI system based on (absolute) pitch tracking. The resynthesizer doesn't have to get the octave right, and latency requirements are usually quite a bit more relaxed too. I speculate this is why Antares quietly removed Autotune's pitch-to-MIDI feature after the first few versions. It has also led many companies developing pitch-tracking products to "pivot" from transcription into resynthesis. Investors get impatient...

[GreyLion]

What's unethical isn't about a bad release, or even a slow one. It's that, more than half a decade later, you keep taking money for vaporware to finance your life, from people who think that iPhones are advanced tech. A quick google-stalk indicates that you have no other source of income. Of course you don't want to let Imitone go, as long as you can get people to pay for you to paint pretty word pictures for them. On your schedule. Which has no endpoint. With moving goalposts.

You keep talking about non-specific 'research' like you have some magical algorithm just around the corner. Allllmost in sight. You go completely dark for long periods of time, then get extremely verbose, briefly, while seeking a new round of financing. With essentially nothing new visible to consumers. Let's quit pretending, after all this time, that the discussion is still about tech, mythical or not. It's about you abusing the trust of your customers, while working hard to bring in new naive ones.

It's hard to not surmise that there will be no algorithm that simulates magic. All the research and coding b.s. aside, for real-time sampling and analysis of the full quick-changing waveform, you need, at the very least, a fast CPU, plenty of RAM, and minimal buffering. (Yeah, let's not bother with the diversionary discussion about rising and falling edges of the cycle, cache memory, die size, and so on. At a certain point, you simply hit the wall.) Maybe you thought that Moore's Law would keep spiraling up. But it didn't. Clock speeds are still generally no more than 5 GHz. And I'm betting that Imitone lives mostly, perhaps entirely, on one core. Yes/No? Got any performance benchmarks on CPU utilization and results with newer CPUs? You can kinda simulate faster ones by increasing buffer size and number, then measuring against lag.

It would probably work great with a 30 GHz clock under it.

TL;DR -- Maybe you should figure out a side-hustle to pay for your Imitone hobby, and quit taking money from innocents until you actually have something to sell.

[Ah_Dziz]

Whatever method waves uses for pitch tracking does actually get the octave correct. In waves tune and the new synth product. Both also do it more accurately than any other program I’ve used.
I can appreciate that you like using the program outside your DAW and that for whatever reason there is no UI for the plugin, but that is not what most of the folks that already paid for your product want.
Hopefully it all gets sorted out in the end

[ThomasHelzle]

If I hear you, Evan, one more time explaining to us how wonderful the external app approach is I think I'll
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAArrrrrrrrghhhhhhhhhh.

In Bitwig it's a mater of seconds to set up the needed tracks, I do it all the time with no trouble at all with dozens of other MIdi VSTs (which I use extensively) and I think we explained to you a hundred times why the external solution sucks for anything but your dog howling into garage band.
Nobody cares if you personally prefer it. Zilch, nada, nüscht, non, zero.
You are the one guy who doesn't count when it comes to the usage of the tool if you sell it. Otherwise call it a hobby and leave us alone with your antics.
BUT DON'T YOU EVER EXPLAIN TO YOUR CUSTOMERS THAT THEY HAVE IT ALL WRONG AGAIN AND AGAIN FOR YEARS.

And that you call us "hardcore types" for trying to do what ACTUALLY WORKS FOR US instead of your personal pet pipe dream is closer to insulting than I think you should go after all this time.

What makes us "hardcore types" is that you withhold a frigging GUI for the VST we bought.
If now you wait for Midi 2.0, all hope is lost, since so far there isn't a single DAW that can actually do Midi 2. It may take years until this is solid.

I have no idea why you seem to grab at the most silly straws for 6 years now to explain to us, why the parts that do NOT NEED YEARS OF R&D aren't there. A VST 2 with Midi out and the same GUI as the standalone could and should have been done about 5+ years ago with the voice to midi tech available then. You could have improved the algorithm from there just as well.
But no, you explain to us that there could be problems with OpenGL (where tons of other plugins do not have any such problems) and that you don't like the idea and tons of other stuff that makes no sense and only shows how clueless you really seem to be.

I mean I've done some procrastination in my life and I've seen people avoiding things before, but I don't think anybody will ever beat you in being as stubborn as if your life depends on it in areas where there just isn't any good reason for it. What's the point in spending all that time arguing against stuff you have to do anyway sooner or later? Just sit down and implement that frigging GUI in the VST2 right now and half the complaints are gone.

I do NOT complain about how long your research takes. If it takes the rest of your life, that's time well spent, you could sell updates and people would be happy to work with you towards better and better tech.
But the totally idiotic thing is not to deliver the easy parts along the way so that people can actually USE your product as it is in the way they chose NOW.
Your standalone GUI does a lot more than the GUI-less VST.
It's a real work of beauty.
I WANT TO FRIGGING USE IT IN MY DAW AS WELL FGS.

Sorry, for the rant, but this is the most ridiculous thing I've ever been involved with.

Memento Mori.

Tom

[paterpeter]

Wow, this still hasn't been released? I'm glad I requested and got a refund when it became clear that this is going down the drain.

Maybe he's run through the $100K that he raised for the Imitone that he couldn't make in five years?

Haha, the only programmer I know who's gullible enough to work 7 years straight on a project like this for 100k is me.

Haha, so funny. People payed you 100k to get a product you advertised as being completed in a couple of months. Haha. Your supporters are really funny - expecting what was promised over and over again instead of financing this ridiculous procrastination-fest. Haha.

[Saffran]

Interesting to see if you succeed in making a hustler care about other people.

[interactopia]

You keep talking about non-specific 'research' like you have some magical algorithm just around the corner. Allllmost in sight. You go completely dark for long periods of time, then get extremely verbose, briefly, while seeking a new round of financing. With essentially nothing new visible to consumers. Let's quit pretending, after all this time, that the discussion is still about tech, mythical or not.

I've written pages and pages about my research process and the results speak for themselves. Virtually every comparison I've seen concludes that our pitch-tracking tech is the best in the world, and 0.10 is the best yet. But it's still hard to use effectively, just like every other pitch-to-MIDI product that has ever existed. At the heart of what we're doing is an unsolved problem (but I encourage others to try).

I'm not some Iron-Man prodigy who can solve a problem like this in a month. Again, it's a technology with a 30-year history of failure and abandonment. My strategy is to continue educating myself, developing my theoretical framework and working at a problem that smarter people have given up on. Do you suppose my customers would be happier if I quit?

You're correct, by the way: with the exception of my work as SoundSelf's engine developer, imitone has been my primary occupation and primary source of income since the Kickstarter. I've chosen to live simply so I can keep working to make a tool that will be useful in the hands of ordinary people.


Tech & Theory

Whatever method waves uses for pitch tracking does actually get the octave correct. In waves tune and the new synth product. Both also do it more accurately than any other program I’ve used.

OK, forgive my nerd rant... Even ordinary telephones use a low-quality pitch-tracking algorithm as part of the way they encode / decode human voice — and it's actually very similar to the voice-driven synths you see in the plugin market. That algorithm will often hop by octaves or intervals like 3/4 or 5/3 of your voice's fundamental — but only when the partial frequencies are sparse enough to create that ambiguity. On the other end, it's reconstructing those same sparse partial frequencies with a formant filter.

Because the formant filter sort of "cancels out" the error in pitch, the pitch detector can get away with all sorts of mistakes. This also applies to intonation correction software like AutoTune and Melodyne (although Melodyne is pretty good at octave resolution).

[...] you need, at the very least, a fast CPU, plenty of RAM, and minimal buffering. [...] Maybe you thought that Moore's Law would keep spiraling up. But it didn't. Clock speeds are still generally no more than 5 GHz. And I'm betting that Imitone lives mostly, perhaps entirely, on one core. Yes/No? Got any performance benchmarks [...]?

Our algorithm runs on one core per voice — I don't know what we'd gain by distributing it. The 0.10 technology is pretty heavy, but only because it is still missing some optimizations. We released it as soon as it was beginning to surpass our 0.9 tech in effectiveness — particularly for staccatos & fast phrasing. The calibration model is still pretty crude, though, and we could be skipping quite a few unnecessary calculations.

I would say that, "luckily", CPUs got fast enough for good realtime pitch tracking a few years before clock speeds leveled out. An ordinary smartphone is sufficient. The limitations now are a matter of information theory — especially at the beginning of a tone. It's like radio astronomy: you can only get so much information out of an ambiguous or incomplete signal, improving with time and SNR. Voice-driven synths get to "cheat" this with transient detection and other methods.

Anyway, we're almost finished with the problem of pitch tracking. Once we can bake a nicer calibration, it should be good enough. However, a truly useful pitch-to-MIDI tool requires another technology entirely: A statistical model of music theory. That will be the focus of our fifth and final phase of research.

I have no idea why you seem to grab at the most silly straws for 6 years now to explain to us, why the parts that do NOT NEED YEARS OF R&D aren't there. A VST 2 with Midi out and the same GUI as the standalone could and should have been done about 5+ years ago with the voice to midi tech available then. You could have improved the algorithm from there just as well.

Fair! Very fair.

Regarding the bizarre, half-baked state of the VST, the truth is that it took a huge amount of time to get it to minimum viable — imitone is built on what used to be a videogame engine. I was burned out by the time I released that, and I didn't get much feedback about it. Even now, only a small fraction of the feedback I get from users is about the VST, and I tend to interpret that as either disinterest or complacency. Far more people ask about mobile, actually... So, my attention turned to other things, particularly participating in MIDI 2.0 (I must be a sucker for projects that take forever).

That said, in its current state imitone is a far more useful tool to pro-audio folks than casual musicians. I acknowledged that when I did the original plugin work, and the team acknowledges it now. A second pass on the VST is on our roadmap for 2020. With the addition of a project manager to oversee priorities and community managers who can shoulder some of our support workload, it should be much more manageable than before. So, I'll take your remarks back to the team.

[GreyLion]

This conversation is a duplicate of the one we had here in 2018, with semi-new justifications, deflections, and feeble techno-bafflegab that fools only the unsophisticated. Read back over this thread from the beginning.

The core is that you took a lot of money, then didn't deliver what you deceptively promised. When you *suddenly* found out that you'd lied to your KS supporters and Imitone wasn't months away from completion, you did not refund immediately until you figured out the parameters of the task ahead. Instead, SIX YEARS LATER, you continue to solicit more money to not do what you're promising. Because, otherwise, you might be forced to search for work that requires actual measurable production.

Every statement you've made this week can be easily and accurately deconstructed as a vague effort to avoid addressing this continuing ethical failure on your part.

I don't think you have any clue at all about how to turn this around and become a legitimate developer. You don't need more research, money, or time. You need self-awareness and to find the ability to listen to and value brutal honesty that pierces your fantasized self-image.

[Radiuss]

The only way to make pitch to midi really work is by instrument/voice-learn recognition. I explain the whole idea behind it here.

What I have noticed with imitone is that if I sing/hum a note in a certain pitch, it creates other midi notes often. Often it is the right pitch, but often not.
Since every voice is different there should be a way to press a C note on the musical keyboard, then press LEARN button, then make a C sound with voice. Every note (in this case C note) should have the option to LEARN 10 different (or multiple) voice sounds all in the C note. Just like you fingerprint unlock on potje phone LEARNS multiple prints of the same finger.
When the C is ready we move on to C#, then to D, then to D# etcetera.
Very few people have such a clear voice that the pitch is (nearly) almost converted perfect by imitone.

Evan, I love your product. I also love to see a feature to learn voice and then translate it to Midi.

I have an idea in mind, a nice melody; starts with an F note. I hum the F note into the mic, imitone converts and sometimes it’s an F note. But a lot of time it’s another note, or a bunch of notes. When I listen back what I hummed (recorded audio) it’s clearly an F note, even in Melodyne the hummed tone is an F note.

If voice/pitch to midi implements a LEARN button, Imitone can learn what is MY F note with my voice F is.

The reason why I mentioned 10 x (or multiple) x amount) LEARN
option per musical note.
I want imitone to recognize my different voice sounds.
-Pooh
-Baah
-Ben
-Taa
-Doo
And so on.

Evan, I hope you can implement this into Imitone so it can work for all users.

Hope to hear from you in this thread.

Cheers, Radiuss

[TabSel]

lol @ Evan

(former, refunded Kickstarter backer, cannot take Evan seriously anymore, having nothing more to say, just wanting to let people know that your money won’t get you what anybody wants you to believe it will get you, please be aware)

[highkoo]

Is there a VST beta?

[interactopia]

Is there a VST beta?

We have VST/VST3 plugins in alpha for Windows. They don't have a proper UI yet, so they're a bit difficult to use (though this depends heavily on your DAW). Practicing with the app will make using the VST a little easier.

We'll be adding our designer interface to the plugins and porting them to Mac as one of our next three updates (the team is still deciding on a priority order). While a majority of our users seem to be more interested in the stand-alone app workflow, plugin work is a contender for top priority and further input is welcomed.

The only way to make pitch to midi really work is by instrument/voice-learn recognition. I explain the whole idea behind it here.

What I have noticed with imitone is that if I sing/hum a note in a certain pitch, it creates other midi notes often. Often it is the right pitch, but often not.
Since every voice is different there should be a way to press a C note on the musical keyboard, then press LEARN button, then make a C sound with voice.
[...]
I have an idea in mind, a nice melody; starts with an F note. I hum the F note into the mic, imitone converts and sometimes it’s an F note. But a lot of time it’s another note, or a bunch of notes. When I listen back what I hummed (recorded audio) it’s clearly an F note, even in Melodyne the hummed tone is an F note.
[...]
Evan, I hope you can implement this into Imitone so it can work for all users.

On-the-fly machine learning is a possibility... but I think supervised training like what you suggest would be far too tedious for the user. I don't think it will be necessary, either... Our pitch tracking technology is already working quite well with a wide spectrum of human voices. The only exceptions I'm aware of are non-modal voice (vocal fry), which has a highly subjective pitch, and voices with false vestibular fold resonance (cord noise or "growl"). We've actually introduced a solution for the second thing in the new 0.10 technology (switch on the atom icon).

Based on what you're saying here it sounds more like you're having trouble with the pitch correction system, which certainly has a ways to go yet. Either that or it could be an issue in the recording setup. Using our in-app feedback system, you can send an audio sample along with your settings. This allows us to study any strange behaviors in our lab tools... When we get one of these we often advise users on how to get better results.

Link: Submitting Audio Samples to imitone support

The core is that you took a lot of money, then didn't deliver what you deceptively promised. When you *suddenly* found out that you'd lied to your KS supporters and Imitone wasn't months away from completion, you did not refund immediately until you figured out the parameters of the task ahead. Instead, SIX YEARS LATER, you continue to solicit more money to not do what you're promising. Because, otherwise, you might be forced to search for work that requires actual measurable production.

When I realized that making imitone into a good product was going to take longer than expected, I communicated with my backers about it, and I listened to what they had to say. I took the temperature. I offered refunds to people who wanted them, but the vast majority of my backers and users have been and continue to be supportive of my long efforts.

Go through the comments on any of my Kickstarter updates, and feel out the ratio. Project creators don't get to curate or edit those comments. Backer revolts have been common for projects that cross their supporters. But that hasn't happened to me. Instead, people are perplexingly supportive as I continue trying to fumble my way to making a new technology. I don't know if I deserve that kind of confidence, but I'm doing my best!

If I achieved that on the basis of communication skills alone, I'm in the wrong line of work. For every hour I've spent writing Kickstarter posts I've spent hundreds on research, development and engineering. I suppose you think that's a lie, though... I don't know what to say. I can only let the results of my work, however incomplete, speak for themselves.

[GreyLion]

*sigh* Your defense mechanisms and web of self-justifications are impenetrable. I'm out for now.

My best advice at this point is for you to go ahead while you still have the money, and buy yourself that nice commemorative watch for when you retire from Imitone Beta Corp. after 20 years.

See you back here in 2022 for the next round.

[TabSel]

lol @ Evan