It doesn't really extract "fundamental" from the audio you import it just adds (pre-defined) resonance(s) to the sound with a very narrow eq boost, and that boost is defined by the amplitude of the audio, that's how I see how it works. It doesn't really use the original audio really for anything expect as kind of envelope follower to the eq boost. Nothing special imho but I guess ok for built with AI. Tough crowd, ey?

 

Last edited: Apr 26, 2026 at 12:48 PM

I did. Although I'm not claiming that I made particularly good and accurate prompts. However I thought that vibe coding meant exactly that - to let the machine do the work. If you start a discussion about the topic, before you know ChatGPT will offer to you "build a basic synthesizer plugin in VST3. Do you want me to do that?" I agreed, but wasn't flattered by how it went on.

I get that you may need another AI for this to work. I heard that people do amazing things with Claude and such, but I've never used them. Maybe I had wrong expectations by just using the free version of ChatGPT, without background knowledge (but a fresh install of Juice and Visual Studio).

My overall view on this is very different however: It's over. These models WILL be able to do all the work in the future in a fraction of the time a human would need. There's no denying - maybe we're just not there yet.

While the majority isn't taking these tools serious right now (maybe myself included) they will amp up their abilities in the background just to surprise everyone with it's power, when it's already too late.

 

haha yah for sure. I think there is some pitch tracking too, or maybe you just set the scale i forget. Def nothing crazy that hasn't been done before. Ive been doing similar with just stock plugins for years. To me its still noteworthy though. Going from zero programming knowledge to a fully working VST instrument, with a fair amount of settings / tweakability is wild. Works as intended, stable, no crazy cpu hit, somewhat unique specialized instrument from nothing is just cool! I think it will be fun to see stuff like this. Someones specific vision on a more niche instrument / effect.

Oh man, ChatGPT is (was?) terrible at this sort of stuff. I have heard that the latest update changes that, but have yet to experiment. Using Claude and creating specialty agents to aid in the creation process you can get somewhere.

 

It was obvious that they would rely on AI to save time and labor costs. But what does the programmer who has to feed this AI with data cost? On the other hand, as with Camel Audio's Alchemy, development times of four years are simply no longer feasible today. I think you could call this development forced rationalization due to the high labor costs.

Initially, however, I believe that AI development is primarily driven by low-wage countries like India, for example. I think in 10-15 years or more, AI will enable the creation of plugins—essentially ready-made modular systems. It will also be interesting to see teams of "crackers" who hack these modular systems and make them accessible to everyone, so that eventually you can design your own compressor.

Anything is possible; it's also possible that many companies will go bankrupt.

 

Thats the thing to get good results u cant just go to chatGPT and give it a simple prompt and copy paste code. That was how to use it a few years ago. But the best way is to get a IDE like Cursor, Google Anti Gravity, Claude Code, OpenAI GPT Codex.

Use any LLM like GPT to help you craft a indepth prompt to create your plugin. Give it instructions on what os you are building for, what framework like JUCE etc.

the LLM can give you a in depth .md (markdown file) of what the IDE needs to do. Then paste that into your IDE it will pull repos, pull JUCE libs, and anything else it needs to build your plugin. It will build debug and deploy for you. Then you can work with the IDE to guide what you want the plugin to do, add features, etc.

 

Hello @curtified, since you are a fan of AI, I and others here would be very happy if you could create an AudioSexPro Supersaw AI synthesizer called AudioSexPro Supersaw v1.0 VST3.

 

You can easily make this supersaw synth yourself! Thats the beauty of understanding these new tools.

They lower the barier of entry so anyone can participate in creation. Since the mundane heavy lifting is being done by the IDE. Its all about how creative you can be on the input (idea) and output (executing the idea).

Maybe try something as simple as a super saw synth to understand how to build the next thing. Just like picking up a guitar and figuring out your favorite bands songs so you can understand how to use a new instrument.

 

Last edited: Apr 30, 2026 at 7:41 PM

I vote for a Pulsewave created supersaw that only works if you ask it a question, and it responds by copy pasting an answer from an LLM. Each letter of the response is a voice for the supersaw. PulseWave, make it happen!

 

One thing to note just like any early stages of advancements tech. Dont look at where things are now. Look at whats out now to see where its going. Soon with local inference on our computers at the speed of inference like what chat jimmy is solving - give it a try it can infer a LLM faster than a google search https://chatjimmy.ai/ also https://groq.com/

We will just be able to communicate with our daw in natural language, to make edits, vibe code "plugins" and other tools instantly. An example is you can use a LLM like gpt to handle pretty much any task you can do in Microsoft Excel. Our DAWs are pretty much Excel for Audio data.

be it clunky right now like most new tech even https://www.producer.ai/ is showing what the future will look like with its tool creation playground.

Or just spitballing here. What does the future look like if we combine isotonik song sketch https://isotonikstudios.com/product/song-sketch-2-pro-pack/?ref=25 with ableton MCP https://github.com/ahujasid/ableton-mcp in a way to get other real time arrangement ideas for a fully produced tracked and recorded song using fully human input?

 

The Deja VU im getting on this thread is of what it was like in the 90s when everyone would look at the tools of the time and make fun of them.

"A computer cant sound better than tape"
"computers wont replace this SSL board"

Yes they were crude and those thoughts were right for the time. But fast forward to now. You go into a major studio and the SSL is just a giant laptop stand and the tape machine is collecting dust in the corner while there is a computer monitor with protools on it in the middle of the SSL board. The old tools are used but the newer ones are used more often.







"This will never be good for pro audio"

 

Historically, this is wrong. Maybe people are missing SSL consoles and tape more right now, because in the 90's everyone fell in love with the idea of no distortion on the storage and lower noise floor at the tracking stage. In fact on the 80's there were plenty of best seller digital tools as synths and reverbs. In fact, most of the OGs with 40+ years of career mixing in the box will prefer clean digital tools if they choose this path.

 

Most of the people I hear waxing poetic online about large format consoles to tape have never actually worked on one.
Artists who have been recorded on them can miss them, because they had the pleasure of recording and listening, but not actually dealing with the grunt work in the control room.

 

I’m saying my opinion or starting a debate about which is better.

That’s not my point.

Im saying when digital audio first came out it would get frowned upon in its early stages. Just like the original post of this thread.

Im the early digital audio workstation era people didn’t account the future of where a MAJORITY of the industry would be using digital tools over the analog tools at the time.

Before you needed to rent out a studio with that gear. Over time anyone can make good music with or without those tools because the barrier of entry has been lowered with new technology. That’s my point here.

I’d argue that a Majority of music today is made entirely in the box. With the external audio touching only an analog to DIgital converter.
There is very little music purely tape and console.

 

Regarding the production stage, i tend to agree, even if i don't like the idea at all. If you open YouTube, most of the new songs have some evidence of AI usage, especially from indie producers.

That's why i think marketing is driving our decisions, for AI and plugins. For AI, i don't think the producers in general are happy with this tech, even if they use it. The end consumer will hate anything that they notice it's AI. The producers are now working more to make legit stuff not to be flagged as AI generated and that sucks.
For the console romance, this is pure nostalgia and the vile marketing saying that if you use the tools from the past you mix is going to sound like ones made on past. And the devs will put a demo song with a perfect tracked 70's style song and use their 70's plugin (lol) to pretend the plugin is making that sound happen, not the arrangement, not the recording stage. And there goes you beloved money.