Loopy Pro: Create music, your way.
What is Loopy Pro? — Loopy Pro is a powerful, flexible, and intuitive live looper, sampler, clip launcher and DAW for iPhone and iPad. At its core, it allows you to record and layer sounds in real-time to create complex musical arrangements. But it doesn’t stop there—Loopy Pro offers advanced tools to customize your workflow, build dynamic performance setups, and create a seamless connection between instruments, effects, and external gear.
Use it for live looping, sequencing, arranging, mixing, and much more. Whether you're a live performer, a producer, or just experimenting with sound, Loopy Pro helps you take control of your creative process.
Download on the App StoreLoopy Pro is your all-in-one musical toolkit. Try it for free today.
Anyone coded a music or other type of app entirely using AI?
Absolutely not posting this as a general love it/hate it discussion on the merits/downfalls of AI (plenty of that elsewhere), but after reading a few posts from developers using AI for other types of software, just wondered how this is affecting music apps and desktop applications. Not from a user perspective (‘make me a funky bassline’), but from the actual creation of the software.
I’ve just read a post where using the latest toolset, a coder simply explained what he wanted the app to do - and AI not just coded it, but launched the app - tested it, and made modifications to the UI and UX based on its own test results.
I notice there’s an app (desktop FX) already where users can customise the UI via AI instructions, but will we reach the point soon where we will be able to design and build our own music making software, just by telling an AI bot what we want?
Comments
Well, it's complicated. With most AI, it's garbage in = garbage out. So it really comes down to how you use the AI. Getting useful information out of an LLM is like watering a flower with a fire hydrant. So you could say "make me a reverb plug-in" and get widely different (mostly bad) results. However, if you have AI in agent mode like Claude Code in Terminal or your development software where it knows the full scope of your project, you feed it the specific documentation, you request small, testable changes, and you know enough coding to catch its errors, then it becomes a very useful tool. There are audio frameworks that are working to improve the output of AI audio coding by providing better agent tooling and configurations like iPlug3, but it is still early. Also, while not discussing the downfalls of AI, I will say the anti-AI coding sentiment online is getting louder.
Right now the tools are best used as a companion, but with more people helping to put guardrails in place, I could see there being a point where people could customize DSP using natural language. It helps that a lot of the data is trained on existing DSP examples, some open-source, but some scraped from the internet without permission.
I'm required to learn the use of AI tools for my job, so as a free time experiment/exercise I'm working with Claude to try to create an AUv3 MIDI plugin framework for Swift, and hopefully some simple MIDI plugins based on that. In my experience Claude is pretty good at generating requirements and boilerplate code and organizing projects, but after a few iterations I still do not have actually working code, and will no doubt have to dive in and start working the code myself to get it to a working state.
I think in its current state, AI is a force multiplier for people who are skilled at coding, but doesn't really allow non-coders to produce a polished app. Seems like it'll get there, but it's not there yet. If I manage to produce a real working framework/app I'll post to this forum no doubt
Thanks!
I’m not a coder as such (just standard web-based stuff), but with only a few years of work left to go (probably just as well, considering) I’m always up for a side-project and learning new skills, particularly if it involves music.
This post grabbed my attention today (I’m focusing on the speed of new development, not the implications of job losses): https://x.com/mattshumer_/status/2021256989876109403
I imagine this could be a bit of a legal nightmare, if it’s grabbing chunks of code from existing software.
Be interesting to see how you get on!
I’ve just been checking out Claude, and it gets mixed reviews. I don’t need AI for my work use yet - it might speed things up a bit - but I’ve created so many templates and libraries over the years it’s all pretty quick anyway.
But definitely interested in brushing up my app building skills when I get some spare time, and a personal music project would be a nice thing to start with.
I'm also working on a non-music-related app with Claude and it's been remarkably quick at getting from concept to a running app. I'm still feeling out what it's good at/useful for. I expect I'll still do a lot of manual work before I get something I'm ready to ship to the App Store, but I've had this app concept bouncing around in my head for literally years, started it a few times and didn't really get very far, and now I have a nearly feature-complete running app after 2-3 hours messing around with Claude. So for an ADHD person like me who has problems getting things started, it's honestly been pretty cool. Of course, the 80/20 rule still applies, so we'll see how it goes as things get more refined and I get into the fine details.
I’ve played around with Xcode and ChatGPT and I can whip up demos really quickly. My ADHD keeps me from getting anything fully finished, and I need to try and make something polished before I can say how good it is from start to finish.
For context, I’m a mathematician/mech. engineer by training, which is where most of my coding experience lies. So I’ve been more of a Python script power user than someone going in the weeds with C. The two proof of concepts I got going are a color palette generator that lets you choose the number of colors to extract from a photo (I call it Color Squash) and a Rhythm Heaven style mini-game batting a ball back and forth on beat. Definitely nothing cutting edge, but I really should finalize that color app since it’s something I’ve wanted for a while for personal use. There are other tools out there but if I build it and understand the original scaffolding I can flesh it out with a lot more ideas.
That scaffolding bit Is probably a smart way to go about it. If it can build you the skeleton of something simple enough for you to understand then but flexible enough to build on you’re approaching it right IMO. A little prior experience goes a long ways. It’s good enough where most of my tech friends could get something working, but my non tech savvy wife wouldn’t be able to prompt anything usable. (There’s a YouTube series idea - wife tries to vibe code)
So for an ADHD person like me who has problems getting things started, it's honestly been pretty cool.
That’s so funny, my ADHD takes me the other direction. I can start 10 projects a day but without a deadline I’ll have 9.5 unfinished projects at any given time.
This is the bit that interests me - I used to have hundreds of ideas for new projects, and in the past some of them have been pretty successful, but they'd take months of trial and error, and I don't really have the time to do that these days. Using AI tools might mean I'll actually be able to get back into a bit of experimenting with some ideas again.
This is the way I'm thinking positively about AI - not how it replaces what I already do, but how it helps me to do things I currently can't - due to lack of knowledge, and time.
I've been lazy over the last few years. I need to catch up!
This is what I need to do - basically just jump in and start something, and then see how it goes from there. I'm more design oriented, so will be relying on AI a bit more for that side of things.
Well, I have that problem too
But I also have the "getting things started" problem. I'm hoping it can help me with the "getting things finished" problem by helping out when I run into a mental roadblock. It has helped in this respect a couple times. My non-music app is maybe orthogonally related to yours; I'm working on an app to come up with color mix recipes for oil paint (apps like this exist already but I've had the idea bouncing around in my head for years and need to get it out).
Hey, that's actually a neat idea. For people who don't know about color theory that could be quite useful. Maybe you take a quick photo of something, use an eyedropper tool and then you get the color mix breakdown.
I’ve updated the title of the thread, as it might be useful to include building apps in general 😎
Sounds like @mjm1138 and I have similar ideas. There’s a few tools out there That do something similar, but a lot of times it’ll pull 1000+ colors from an image (if not more). I usually want to get palettes for pixel art, so having something to condense an image with thousands of different shades into one with 16 is something I’ve been wanting.
Again, it’s a basic idea but I’ve got some ideas of what to do with it from there.
About the color part, maybe you’ll find this interesting https://iquilezles.org/articles/palettes/
Yeah, that’s the idea. I’m pre-populating it with some beginner-friendly oil paint sets, that can be organized into a collection of palettes. A stretch goal would be taking a photo and having the app generate a palette for you from your stock of paints based on the photo, which sounds a bit like the idea @FizzyLizzy27 describes.
That’s pretty cool!
I’ve built small audio tools by letting AI draft rough code, then I clean it up myself. It speeds up experiments, but I still handle structure, UI, and debugging.
I’ve made a pretty comprehensive random walk midi generator VST3 with Juce. But I won’t be releasing it, since the UI is incredibly bare bones (ugly) and I regularly add new functionality that breaks previous presets and occasionally screws other functions.
My prediction is that by this time next year, anyone on this forum will be able to create a professional-quality iOS music app with AI.
Feel free to come back in a year and decide if I'm a prophet or completely insane.
Actually the year after next I use AI to build a Time Machine so I’ve just come back here to tell you that you are correct 👍
I've been using Cursor and v0 to build a browser based app in VS for the construction company I work for. Construction technology has historically lagged way behind the times, so i'm taking this as an opportunity to learn and create something on my companies dime. Honestly, the few people in the industry I've shown it to have been blown away.
What I can say as someone who prior to a couple of months ago had zero coding experience (outside of Visual Basic in Excel) is that your chances of successfully creating something useful or useable solely relying on prompting AI is pretty small. If you are creating something remotely complex, the tech just isn't there yet. I've found sometimes it's like speaking to a child where you can't simply say "create this functionality", buy you have to spell out each step "do this, then this, make sure it's connected to this, etc." if you want decent results. Even then you're likely to break something else. It's all about planning and creating a roadmap.
Since my company is paying for the AI suite I'm using, I've taken a couple of opportunities to test it out creating a browser based sample player trying to merge the functionality of Flexi Sampler from Drambo and the ease of sample editing in Koala. The results have been meh and it's not something I would ever release to the public for use. I fully believe that unlike construction where I have over 20 years of experience, my lack of experience with music apps/software, terminology, etc. is a limiting factor. I will say the design of the app instantly looked like an AI coded app (surprise lol). I could probably get the app to function as intended but it would take away from the limited amount of time I currently have to invest in making actual music. I'll stick with learning about and creating music with apps created by folks with far more talent and time than me.
Bottom line is that AI can do some cool stuff but the notion that you can sit down and blindly create an app is maybe more simplistic than the actual reality of it.
My prediction is that by this time next year, the people developing this technology will realise they're doing it wrong.
Programming languages were designed so that humans could develop software more easily. I don't think AI software development tools will always generate human-readable code.
Right now of course, it's good that they do, because sometimes that code is complete garbage...
But, having the machines generate code, that less and less people are willing / able to understand, is fundamentally flawed imho.
For Augmatic GRE, the developer has claimed that it is the "first 100% vibe coded plugin":
https://forum.loopypro.com/discussion/comment/1458092/#Comment_1458092
As far as I can tell, It seems to be widely considered as a quality app, as good as or better than many professionally-developed non-vibe-coded apps.
I will just point out that, if human-readable, highly editable code is a priority, you can steer AI in that direction. I asked ChatGPT how to do that, and it suggested including the following prompt among the directions. I believe there are already add-ons to vibe-coding environments that do something similar, and expect that they'll get better in the future, as AI continues to progress:
ChatGPT said to include these instructions:
Hey, coming back, I ended up making and publishing my idea.
https://apps.apple.com/us/app/colorcondenser/id6760206595
I should probably drop the price a bit, and I have a few UI and advanced feature bugs to work out (video import is a bit unstable with long videos). But I personally use it a lot day-to-day to get reduced palettes. I might’ve priced it a bit high.
I’ve got a new baby and a new MacBook, with not much childcare until recently so a lot of work got put on hold.
Also, I’ve been experimenting with local models building small personal apps. I’ve made a few little utilities I run natively using local models, including a tower defense game with some fun mechanics. It’s pretty impressive what local models can do, and the less reliance on companies who change models on a whim is a plus.
If devs are really going 100% ai coded, agents are pretty good at looking through a code base and piecing together what’s going on. Especially if you use good agent.md, spec, and readme files. Keeping good documentation goes a long way, whether for human or ai dev work.
I think a lot of issues with AI coding is less AI capabilities than a communication problem. If (royal) you, with no real clue what you want, hired a human developer and said to them 'make me a sampler' I guarantee you the result won't be much better (probably worse, certainly MUCH more expensive) than what AI will generate in a few mins. The questions is, do you care enough to then see what came out and iterate on it? and again? and again? until eventually it's something? Surely this is what the manual process has been, why expect that part of it to go away just because you don't know how to code? Regardless of your experience with code, if you have no proper vision or the motivation to actually keep iterating until it's any good you will fail, imho.
BTW, I vibecoded a project to help with vibecoding (primarily iOS audio apps but you can do anything), and have used it in my own processes to communicate better with AI. Essentially it's an intentionally 'ugly' (fast) frameworking tool specifically tailored to creating AI readable JSON.
The real issue is scalability and long term maintenance for anything mildly advanced.
An AI could undoubtedly create the various blocks that comprise a DAW, but keeping the big picture and having code discipline, that simply won’t happen without human oversight. At least for now.
When everything is 100% vibecoded, the AI often rewrites entire swaths of otherwise fine code and creates new problems in the process.
For example, in your new DAW, when users point out that bending notes when multi selected in the timeline causes odd behavior with the UI - good luck fixing that without breaking something else.
Simple from a to b sound processing is not a problem, but how do you really know if the filter algorithm you asked for is actually the one implemented? And is the AI making “helpful” decisions in the background about how the sound should mix or perhaps something you didn’t think to specify explicitly? Nobody knows until someone reads the code.
I'm pretty sure that, right now, when someone says "100% vibe-coded", they're usually referring to a situation where 100% of the code (or very close to it) is written by the AI. I don't think 100% vibe-coded means there is no human oversight. Actually, the opposite, since if you're not going to tweak the code you're going to be especially careful to provide oversight, and request tweaks when something in the app or the code is not to your liking.
Just wanted to add one comment — my plugin AudioPipeline was completely made just using Claude Code 🙈
I did not write a single line of code. I deeply understand what it is doing and how it is implemented, it was me who made most of the architectural decisions and I really deeply managed Claude on HOW to write, to the level where we discussed what parts of code should be optimised, what should be made in Swift, what in C, where to use the Accelerate framework, how to optimise code, etc. — but yeah, in plain terms of typing code with hands, it's all Claude Code. I did not write a single line of code. It took me 5–8 days (approx. 4 hours per day) to fine-tune it to the current state.
Disclaimer: Coding is my job, I have 30+ years of coding experience in developing large-scale applications, frameworks, database systems, etc. I have serious knowledge about DSP coding and a partial understanding of the math used for DSP code. So I am not a typical "I don't know what I am doing" vibe coder
)
I don't think people without deep knowledge and MANY years of coding experience will be able to make a complex app just using AI anytime soon.
Not in one year. Not in two years. It's not going to happen.
Why? Without a really deep understanding of what the code is doing, in more complex apps you end up with an absolutely messy code structure, with duplicated code multiple times over — exponentially harder to maintain and more and more fragile, where a change in one place causes an entire new group of bugs in a completely different place.
It's EXTREMELY complicated to manage this thing to write consistent, well-structured, maintainable code. Not impossible. But extremely complicated — it needs a lot of patience and a lot of knowledge about how code should be properly structured. Something that people without coding knowledge simply don't have.