Martinic released Doctor Mix AI Synth, a free VST plugin made with ChatGPT.
It’s hard to escape the constant deluge of AI-related news these days. We recently featured Emergent Drums, a drum synthesizer that uses AI to generate new drum sounds, and the amazing Enhance Speech vocal enhancer from Adobe.
More AI-related news has emerged as Martinic introduces the Doctor Mix AI Synth plugin. Instead of using AI to generate sounds, this free synthesizer was developed using AI. It was made with ChatGPT and is a fascinating proof of concept of things to come.
If you’re interested in examining the VST code generated by ChatGPT, you’re in luck because Doctor Mix AI Synth is now open-source (available on GitHub).
Doctor Mix AI Synth is a monophonic subtractive synth by design. Whether you’ll like its interface or not is a matter of taste. You probably won’t mind the design if you’re a fan of Doctor Mix since you’ll be seeing Claudio Passavanti’s face as the most prominent element of the UI.
There are a handful of controls, as the synth is limited in scope. Oscillator waveforms are restricted to a saw wave, with no option for other waveforms selectable. The saw wave doesn’t seem to suffer from aliasing, but it’s difficult to tell since there isn’t really much in the way of modulation or syncing available on the oscillator.
A single filter and envelope round out the rest of the controls. The envelope is disabled by default and functions as the amplitude ADSR when toggled on. Filtering has minimal controls, with the cutoff, resonance, and a fixed LFO with rate and depth selectable. I didn’t test the waveform of the LFO, but it sounds like a sine wave when active on the filter.
Doctor Mix AI Synth isn’t going to ignite the world of synthesis. However, it serves as an interesting AI experiment that came as a result of a collab between Martinic and Doctor Mix. The synth is, of course, free to download, and early downloads get a 50% off coupon for the more full-featured AX73 from Martinic.
At the very least, it makes for a fun diversion with an arpeggiator, and I had a good amount of driving with free EON Arp from Modalics that we covered just a while ago.
Doctor Mix AI Synth is available for Windows and Mac computers, with available formats being VST, CLAP, and AU. There is no indication from Martinic as to whether it is available for Apple Silicon devices natively.
What are your thoughts about this project? Do you think we’ll see more AI-generated VST plugins in 2023? Let us know what you think in the comments below.
Download: Doctor Mix AI Synth
More:
13 Comments
MRG
onSo, it’s MeowSynth but instead of a cute kitteh, we get …that guy’s face… mmm.
Joking aside, it’s scary/amazing/awesome the results you get from AI, considering that: 1. this is only the beginning of the start of a decent grasp on what will be possible, and 2. that ChatGPT isn’t even designed to do coding, but does it on a conversational framework. What will a specialized CodeGPT be able to do?
Yorgui Hartmann
onyou dont need advanced ai to make this tbh.
anyone with a synthmaker kit can do that: the most effort is likely been done on the gui. which in this case is a joke haha.
Numanoid
onI am not very much impressed with electronic music of this day and age.
With the technology at hand we should explore beyond 12 tone music, and go deep into micro pitches, as well focusing on everything dynamic.
Yet the hits played on radio are just in regular major or minor, three chord songs, like 60 years ago, and the main thing is to make it sound as loud as possible.
And that is because that is what people at large want, so it is impossible to win, if you want to make a living out of it.
MRG
onIf you’re using microtuning in a ‘normal’ fashion, people think it’s out of tune. Or you can go nuts and it just sounds weird. :D
The lack of interfaces to play with them makes it harder to explore, I guess.
Numanoid
onSlightly out of tune is what makes Boards Of Canada, such great listening IMO :-)
But to employ twelve tones is even more important, there is far too little tritone in modern popular music.
MRG
onIf you think about it, music was always not perfectly in tune. Only perhaps the last 25 years who kept everything digital, used autotuners, etc… got really ‘perfect’. I do like being out of tune, I might often overdo it. :)
Couple years ago I had some fun with 17-tone, I liked it for being a nice middle ground between sounding detuned and not too microtonal mayhem, perhaps I should go back to it for some more.
Santana
onFoi comprovado no Brasil que por traz do chatgpt exite manipuladores democrático, que prestigia um lado eleitoral e genérico 😬
MRG
onYou have to remember that AI is basically only simulating a response, based on what -we- are feeding it. If you want AI to be more human, perhaps we should be more human ourselves first? (You have two hours, and four sheets of paper). And play some Mônica no Castelo do Dragão.
Dayz
onI don’t believe someone just told ChatGPT a few instructions and it outputted a working compiled vst. Like where is the bts and video of building this?
It is requiring many and much understanding how to instruct it.
MRG
onThe BTS video is literally on the webpage.
I would recommend watching two videos from “Burned Guitarist” on YouTube too, named “Can ChatGPT Build A Guitar Pedal Plugin?” and “Can ChatGPT Build Guitar Amp Profiler?”.
MG
onNext level douchery…
Tim
onHow would a text predictor make music?? Code presumably can’t be written this way with any sort of understanding of the musical output/quality. At best I could imagine its a randomizer?
MRG
onI mean, it’s rather simple. ChatGPT was fed various audio frameworks and coding languages rules, it probably spewed out the code from the samples that came with the Steinberg VST3 SDK… :p
So yeah, don’t expect revolutionary code from this, as of now. In the future, perhaps.