I made some synths & audio processors.
Hey all!
These are a couple of ideas for synths/plugins/effects that I wanted to try out with ChatGPT first as like a proof-of-concept before going to any plugin company to maybe turn them into a reality. They run inside your browser, some of them require you to load a .wav file into them.
I cannot code so I had to rely on our AI overlords here, sorry for that, but I thought maybe you can get some use out of these, too!
I might make a YouTube video about these and how to make them and how they work, but wanted to share them here first asap :)
Let me know what you think \^-^/
Comments (31)
My heavily pondered opinion is this: I have no problem AT ALL with letting ai turn knobs. I have absolutely no problem with building gadgets with do real synthesis with AI, including concat devices to an extent which is different because it runs local and depends entirely on what you feed it. I love the idea of a world where artists have their own proprietary tools. I do instinctively know that generated audio / Video may be harmful in ways we don't yet understand. I know where my line is. This got my stoked to see VR is applying himself in AI because his results are going to be insane.
Hi Virtual Riot, I think your first Synth has some serious Sound design possibilities and potential. I love the innovative thinking when it comes to all things bass and sound design. Don't sleep on the 1st one you showcased in your recent live stream, (What if per detune voice somehow had it's own Warp mode like Sync, bend +/- and ect).Also if the real plugin developers could help you with this, could resynthesis on your oscillators be possible like in Serum 🤔 . Your sample based ones also have serious potential for raw up-front sound design and texture manipulation. Love the innovative thinking man, don't doubt yourself 💪🔥🧠. It's time for Sound design and innovation to reach new heights like never thought of before 💯📈
Your ideas/projects are awesome - I vibe coded Combfield - MacIntel Vst3
Could you record only waveforms in .wav from analog synthesizer you have and make it as a pack
TLDR: A lot of us have a clear line in the sand in using AI to generate music or visual art. Why don’t we have the same for other AI use? A lot of the same qualms folks have with AIs like Suno apply to all other AIs. I challenge everyone to find that line in the sand for themselves when it comes to AI coding. I know there’s been a variety of opinions expressed here. I’ll speak to my experience as an audio dev. Yes, I understand the accessibility, convenience, possibilities vibe coding provides, but I want to provide pushback on the cost. I see a lot of people assuming the ends justify the means, but never really investigate for themselves where the line in the sand is with their personal use. There’s the obvious - the environmental cost and justification for building data centers. There’s others - this tooling is only able to work because it was trained in the work of countless developers’ code without their consent, credit, or compensation. There’s also the potential threat of replacement of human career paths. Rampant usage of AI contributes to the overwhelming manufactured consent to these AI companies to continue their practices unchallenged - building data centers, scraping the internet, etc. I’m not trying to assign blame or shame, but want to make these tradeoffs clear. Yes, AI has been integrated into a lot of dev’s workflows. Yes, it’s a service that can be used to increase productivity. However, it’s used in varying degrees - there’s a difference between vibe coding and more intentional use that most devs use. Vibe coding leads to inefficiencies, bad architecture, and spaghetti code which is not maintainable. In my experience, AI is helpful with documentation, writing tests, and changes small in scope. Heavy reliance on it results in the issues I outlined above as well as deskilling. For that reason, if you had provided me the source code for these prototypes to port into JUCE or whatever, I don’t know how much I’d actually read and reuse. All that being said, me, I’d much prefer something built in Max/plugdata/phaseplant/bitwig/ableton racks/etc. however, that’s just me. All in all, I just challenge everyone to approach AI usage with the costs in mind and determine what level and type of usage is acceptable given these costs.
My opinion: I’m not cool with your usage of AI here. I get the function it provides, but there are alternatives that don’t have the same baggage.
Ive not used ChatGPT myself, but what Claude can do if you're willing to be an alpha tester on the plugins you ask it to make is... concerning. But if tech like that is out there, people will use it, which puts you not using it at a major disadvantage. It can code full on VSTs and design really good GUIs, especially if you ask it to make the GUIs good. The thing that sucks about it is just that plugin devs probably can't compete with that, I suppose unless they use it themselves but perhaps for them it doesn't make their jobs that much easier im not sure.
SliceSort is really really cool, gives really interesting sort of sounds with a lot of sound design possibilities
Are these compatible either fl studio or only abelton?
hell yeah bro, love that you are using AI in a way that helps you forward your creative vision, love to see it. can't wait to check these out!
https://tenor.com/en-GB/view/looking-at-paper-guy-looking-at-paper-jimmy-butler-its-so-peak-it%27s-just-so-peak-gif-14820398212728935760
in the combfield synth when turining knobs i can hear some noise, is it supposed to be that?
Could be some artifacts, if this gets turned into a proper VST that will be fixed :)
Yooo val this is so trippy, as you were making your "chromasynth" i was making https://fractal-beats.lovable.app/ something called "chromachord" lol ai obs likes the chroma, but weirder than that it sounds like were making the same idea at the same time lol how weird is that. Mine is more aimed at making the visuals though, i aim to make a music video maker, using similar randomization features used with abletons macros except to make music video visualizers and you can feed it your song and stuff. Ive been wanting to make vsts but couldnt find a decent place to build them, i guess just gpt was the way lol
:-(
is it that much better to use ai than find someone who can code :\
I code for a living, and the majority of us aren't hand writing code anymore, tbh, even audio devs
Still considering a Super Slicer plugin? I can imagine throwing my mudpies into the greatest happy accident machine af all time!
Should be done next month :)
thank you !! very cool
Check your IG DMs
Admittedly I am someone who struggles with the normalzation of the use of AI but i'm glad you are finding fun ways to use it! I look forward to using these tools and will report back!
Just bounced a lot of stuff with the SliceSort one and just realized I was using a already processed file. I was wondering, if it's possible to get the same result again with the unprocessed file via the Seed code maybe. I tried typing in the Seed again but that doesn't work. Is there any way to recreate the exact same pattern after replacing the WAV ?
Im a music software developer, would be super down to work with you for free as a partner. I also love sound design and audio engineering so reach out to me at elon@chosenmasters.com
as long as it works 🙏🙏🙏
sick! ok, im back with my verdict. i love combfield the most, but i also really like fanout-synth and spectral darkroom is pretty cool too!
I'm not sure about ChatGPT, but Claude is very good at working with JUCE (the framework virtually every VST uses) to create actual VSTs. Just FYI if you had any interest in pulling them into a DAW to play around with. it'd work especially well given that you already have working prototypes for it to work from. Just tell it you want it as a VST and a standalone app.
this is super intesting. ty val!
Honestly, AI being used for Open Source or proof of concept is in a lot of people's opinion perfectly respectable. It's the people that use AI and claim it's their own that lose respect. Just look at the Decomp Recomp scene for retro games, people are finding respectable uses for AI and POCs are one of them!
Can you route these into Ableton somehow via drivers or are you essentially relying on sampling while using these?
Im not entirely sure i understand it but thank you Val!