yes of course, you nailed the issue.

so far about our "why".

not to the "why" of those who are in control.

the industry knows about that development, too.

they are speculating that his happens. and then streaming will die, but the prices to use those closed generative models to create you own music on the fly will rise.

 

A friend here on audiosex asked me why I don’t reply to some idiot who writes offensive comments…
I replied with a quote from Oscar Wilde: «Never argue with an idiot: he’ll drag you down to his level and beat you with his experience.»

If a client buys one of my songs but then asks me he/she want to present that song to their record label with an AI mix… I try to work out how to do it and do my best to accommodate them.

I’m making money with audio AI; of course I was making money before too, but now I’m making more.
And for now, nothing’s going to make me change the way I work.

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

I’d like to thank everyone for the tips and suggested audio experiments to achieve ‘that’ particular sound… (I wasn’t asking for opinions on which sound was best or whether you liked AI or not… that wasn’t the point). My album was mixed on an SSL 4000 E… so what?

I’m trying out various suggestions you’ve given, and I’ll go with the one that will works best.

 

Last edited: Jul 17, 2026 at 2:19 PM

So you couldn't use any of these plugins without wrecking your mix? Uh huh ok then.

 

ElMoreno, at least you're honest. Money isn't art or beauty, but it's a great motivator.

Ayn Rand famously stated in The Fountainhead: "To sell your soul is the easiest thing in the world. That's what everybody does every hour of his life. If I asked you to keep your soul - would you understand why that's much harder?"

Two questions though, as I can't get my head around why a "client" would specifically ask for an AI mix?

Also, why didn't you just go straight to source and ask AI for the answer to your initial query?

 

Hi Colin,
I’ve got a couple of minutes’ break, so I’m happy to reply.

I’m afraid I’ll have to disappoint you, though…
I haven’t sold my soul for money; money can sometimes be just a consequence of what you enjoy doing.

You might be surprised, but I absolutely love working with AI; I’m truly passionate about it. In a way, it has inspired me, giving me back that sense of adventure that I’d perhaps lost a little, stuck in the usual routine of writing (I also write books).
I’ve managed to create a super-efficient ‘model’ that gives me excellent results.

Ask the AI for the answer?
That’s the first thing I did. Aha ha aha haha hahaha…
The answer was complex – more than 30 plugins and pieces of hardware involved in the operation (logically, I imagine, via super super-fast computers) – but, at the end, without suggesting any settings… and that brings us back to the beginning of the story.

As for your first question, however: it’s a much more mundane matter.
If you go back and read some of my posts from a few years ago (many laughed or took offence, as idiots usually do), I’d already predicted what’s happening… These days, people listen to music on their mobile phones, or at most via an iPad or computer… They’ve become accustomed to mediocre sound quality. Perhaps AI is ‘standardising’ a sound texture – a sound – which, whether you like it or not, works well precisely through those mediums.

Whilst the industry tries to sell hardware and software to those who believe that music is made using 1,000,000 kHz / 1,000,000 bits and fleeces them for money, money and more money… the public listens to 128 kbps MP3s and thinks they’re brilliant.
In some shows I’ve sung on songs at 128 kbps (just using a little EQ) and people liked them much more than others of very high quality (it’s the song and the singer that matter… the rest doesn’t count for anything).

Then everyone’s free to think what they like… that’s fine anyway.

 

Last edited: Jul 17, 2026 at 4:03 PM

Don't be shy to use OTT/Boost on every channel, personally think Boost is more flexible and you can really dial down and customize it without things sounding over the top, I have been using it in small doses even on ambient dynamic mixes, just hair of it in parallel goes long way.

 

Thanks, and an interesting observation regarding quality. The counterpoint to that might be that AI training data has been largely scraped from low quality sources. Regardless, your point is valid all the same.

 

I don't think this logic applies with asmr porn, onlyfans and all the stuff like that, believe it or not some people aren't narcissistic that much to demand their own self-proclaimed content

 

lol.

or "how did you mix music during the last 30 years if you dont know how to turn the knobs on your compressor plug-in?"

 

This is what GPT suggested when asked, and not that much of it is actually just limiting and compression.

Spoiler

It is actually a more interesting question than it sounds, because "AI sound" is not really a single processing effect. You cannot take a mastered WAV, run it through an "AI-ify" plugin, and reliably make it sound like it was generated by AI. But you can imitate many of the artifacts people associate with AI-generated music.

The trick is to remove some of the things that make a human mix feel intentional and introduce the kinds of imperfections AI systems often produce.

Some approaches:

1. Destroy the normal mix balance
Commercial mixes usually have:

• clear vocal hierarchy
• controlled low end
• intentional stereo placement
• consistent dynamics

AI-ish mixes often have strange priorities:

• vocals too loud or too buried
• instruments fighting in the same frequency range
• random elements appearing too prominently

You could:

• boost some midrange instruments 3–6 dB
• push vocals forward unnaturally
• narrow/widen random elements
• make the kick and bass relationship less coherent

2. Add "latent reconstruction" artifacts
A lot of AI audio generation has a slightly smeared, reconstructed quality.

Try:

• heavy spectral denoising followed by reverb
• extreme transient smoothing
• multiband saturation
• low-level granular processing
• subtle chorus/flanging

For example:


WAV
↓
RX Spectral De-noise (too aggressive)
↓
Granular processor
↓
Very short shimmer reverb
↓
Limiter

You get that "it was rebuilt from a description" feeling.

3. Make the timing less human
AI generations often have:

• slightly unstable groove
• odd fills
• inconsistent microtiming

You could:

• slice drums and randomly shift some hits by 5–20 ms
• apply random MIDI-style velocity changes (if you have stems)
• use tape wow/flutter very lightly
• automate tempo ±0.5 BPM

4. Reduce harmonic realism
Many AI songs have strange harmonic behavior.

You can mimic this with:

• pitch shifting small sections by cents
• formant shifting vocals
• weird autotune settings
• spectral freezing
• resynthesis effects

Tools like:

• iZotope VocalSynth
• Soundtoys Little AlterBoy
• Output Portal
• Arturia Efx Fragments
• Melodyne with exaggerated pitch/formant changes

can help.

5. Make the mastering "too perfect"
A lot of AI music has a recognizable polished-but-flat sound:

• very compressed
• limited dynamics
• bright top end
• slightly blurred transients

A chain might be:


Multiband compressor
↓
Exciter
↓
Stereo widener
↓
Aggressive limiter
↓
Tiny bit of clipping

Basically "streaming algorithm demo track" mastering.

6. The closest method: actually re-generate parts
If you had stems, the most convincing way would be:

• separate stems with Demucs/UVR/Spleeter
• run vocals/instruments through AI transformations
• regenerate sections with AI models
• recombine

A whole-song WAV is already baked together, so you are fighting the mix.

The funny thing is that many "AI artifacts" are actually the opposite of good mixing practice. A good human mix is about coherence; AI-ish sound often comes from slightly inconsistent coherence.

A human engineer might spend hours fixing exactly the things you would deliberately introduce to make it sound AI-generated.

It's ratting itself out.

 

Here's the definitive suggestion for "mixing like "AI"". Get some cotton balls - slap a cotton ball in each ear - put on your cans over the cotton balls and turn the gain way up. NOW you're ready to "mix like "AI""...

silly wabbit..

 

BWAHAHAHAHAHAHA!!

 

Did you see the very last episode ever of the Loony Tunes production, the one when Elmer J Fudd wanted to make a how to hunt video and needed a film crew? Bugs came up with the idea to use a compressor to lower the sound of the rifle shot during post of Fudd's voice over. He coined the phrase "ducking" but Daffy thought that it was trick to get Elmer to shoot him and so tried to recoin the technique "rabbiting". It backfired on him of course, as when Elmer became frustrated with the two's antics drew his riffle to pop them, Bugs called out "duck" but Daffy instead of getting out of the way was shot through the heart, and just before he died the last word out of his mouth was..."Rabitt".