If you would rather keep AI out of how you make music, good news: most of what you already own is fine. The trick is knowing which tools genuinely use it, which ones only wear the badge, and how to tell the two apart.
For anyone who wants the sound to be theirs, with no model and no training data anywhere in the chain.
Worth saying up front, because the panic is overblown. A subtractive synth is oscillators and filters. An EQ is maths on frequencies. A compressor watches a level and turns it down. A convolution reverb multiplies your signal by a recording of a room. None of that is AI, none of it ever was, and none of it suddenly became AI because the word got popular.
So the plain version: the overwhelming majority of synths, samplers, EQs, compressors, reverbs, delays and saturators are ordinary digital signal processing. If that is your whole setup, you are already where you wanted to be.
It clusters in a few specific jobs, and they are usually the ones sold on the promise of doing the thinking for you.
| Job | Often uses a trained model |
|---|---|
| Making a whole song from a text prompt | Yes, this is the core AI-music case |
| Stem separation, pulling vocals out of a mix | Yes, almost always a trained model |
| "Assistant" mixing and mastering that sets a chain for you | Usually, to some degree |
| Smart EQ that decides the moves itself | Sometimes, though some are plain analysis |
| Vocal tuning, de-noise, transcription | Increasingly, under the hood |
| Ordinary synths, EQ, comp, reverb, delay | No |
Notice the pattern. The AI turns up where a tool claims to make a creative or corrective decision for you. Where a tool just processes the sound you give it, there is nothing to train and nothing trained.
Here is the annoying part. "AI" sells right now, so a fair number of products have an AI sticker on the box and old fashioned code inside. An arpeggiator becomes an "AI melody engine". A preset picker becomes "AI powered". It is marketing, and it works, and it muddies the water for anyone actually trying to avoid the real thing.
It cuts the other way too. A few genuine AI tools get sold softly as "generative" or "smart", which sounds tamer than it is. If you care about this, the badge on the box is not the answer. The question underneath is.
Four questions get you there almost every time.
Trained models are large and often need a server. If a plugin is a few megabytes and runs happily with the wifi off, there is no model in it.
This is the giveaway. AI tools describe a dataset, "trained on thousands of tracks", "learned from". A DSP tool describes a method, "a ladder filter", "a convolution of real rooms". Read the manual, not the landing page.
Ask, or read closely. If the maker can tell you the actual rule behind what it did, it is code a person wrote. If the honest answer is "the model decided", it is AI.
A generative but non-AI tool varies on purpose, within rules you can see. An AI tool produces a new prediction because that is what a model does. This one is subtler, but combined with the other three it settles most cases.
"AI-assisted" and "AI-generated" are not the same claim. A synth with an optional AI preset finder is still just a synth when you never press that button. What matters is whether the actual sound coming out was generated by a model, or by you and the instrument.
This trips people up, so it is worth being clear. A tool can compose and play by itself, endlessly, and still have no AI in it at all. Music theory, a bit of controlled randomness, some synthesis. Rules a person wrote, producing more than you put in. That is generative in the older sense, and it predates the current wave by decades. We wrote a whole piece on the difference in generative music vs AI music if you want the long version.
ARGISH, SILT and REHEAT compose and perform on their own, but they do it with music theory, synthesis and DSP. No trained model, no dataset of other people's music, no AI generated audio in any of them. They are small, they run offline, and the logic behind what they play is written down, not learned. You set a key and a mood, and what comes out is yours to release with nothing hanging over it.
You can hear any of them free in the browser, nothing to install.
We write these when there is something worth writing down. One email when a new one lands or a new Tunary instrument ships. No newsletter, no schedule.