Mixing Without Faders — Why We Built a Chat-Based Mixing Service

More than hear it — I've seen it up close. Trackmakers with genuinely great songs sitting on their hard drives because they can't nail the mix. There are so many of them.

Mixing is a completely different skill from songwriting. Writing a song is about deciding what to say. Mixing is about adjusting how it sounds. It's like asking a painter to pick the frame. Not impossible, but not necessarily their strength.

That's where the problem is.

There's a price cliff in mix outsourcing, right? On one end, you've got people on platforms like Coconala (a Japanese freelance marketplace) or Fiverr doing it for tens of dollars. On the other end, a professional mixing engineer costs an order of magnitude more.

Tens of dollars or hundreds of dollars. Nothing in between. And the problem with the cheap end? Quality is a total gamble. You don't know what you're getting until it comes back. And if the result is mediocre, it's hard to say "hey, this isn't what I wanted at all."

Because the other person is human, right? You can't keep telling someone who took the job for a few bucks to "redo it" over and over. Revisions get awkward. I think this is the most overlooked problem in mix outsourcing.

Exactly. Mixing is supposed to be an iterative process. Even in pop or commercial production, you go back and forth — "push the vocal forward a bit," "bring out more kick definition" — multiple rounds. But that communication cost doesn't work at the budget tier.

Yeah. MUEDial's concept is dead simple: mix by chatting. No faders, no sliders, nothing. You just type "bring the vocal more forward" and that's it.

Does someone outsourcing their mix want to touch sliders? Of course not. They just want to communicate what they want. The reason past mixing services had sliders is because they were built from an engineer's perspective.

The user just wants to say "make it a bit more sparkly, you know?" And translating that into technical parameters — that's the AI's job.

As many as you want. Zero awkwardness. "Nah, that's not it, one more time" — with AI, there's no guilt whatsoever. And every time you ask, you're getting better at articulating your own taste.

Right. After saying "I want more low end" a few times, you realize, "oh, I guess I like bass-heavy mixes." Self-discovery through mixing. It's the same structure as MUEDnote's production logs, you know?

$8.

Budget haircut chains didn't kill salons, right? I think the same thing will happen here.

People doing mixes for tens of dollars on freelance platforms — honestly, I think they should stop. If that price point is burning people out, let AI handle it. Humans should spend their time on things only humans can do.

Not at all. What a professional mixing engineer does is a continuous stream of high-level decisions. Understanding the context of a song, reading the artist's intent, making musical judgments. That's a completely different world from a $8 AI mix.

Exactly. Too hard to do yourself, too expensive for a pro. We fill that gap.

We use RoEx's engine.

Nope. A mixing engine isn't about implementing EQ and compression individually — it requires integrated decisions about inter-track balance, spatial processing, the whole picture. MUEDear's DSP was about applying one process accurately. MUEDial is about how the entire thing sounds together. Totally different dimension.

You don't have to build everything yourself. MUEDear's DSP needed to be in-house because it directly affects ear training accuracy. For MUEDial's mixing engine, great solutions already exist. So it makes way more sense to ride on that and focus our energy on UI and UX.

In the MUEDear article I said "know the landscape, then deliberately build." This time it's "know the landscape, then deliberately don't build." Same decision axis — what's better for the user.

That's the thing. Mixing ultimately comes down to whether you have a reference point for "this sound should sound like this." That reference comes from the accumulated memory of listening to thousands of songs.

It's learned from a massive amount of audio and makes judgments like "for this combination, this balance works." Same mechanism. But the critical difference? AI has no taste.

Judgments like "this direction suits this artist" or "this mood calls for this atmosphere" — current AI still can't do that well. So you tell it through chat. The user's words become a stand-in for the AI's missing taste.

Right. And that's why MUEDear exists.

If your ears are trained, you can say something specific like "clear up the vocal around 2kHz." If they're not, all you can say is "it sounds kind of muddy." Train your ears with MUEDear, put them to work with MUEDial. They're connected.

Ultimately, democratizing mixing. Using AI to close the gap between the ability to write songs and the ability to mix them.

I don't think professional engineers will become unnecessary. But there will definitely be fewer people who can't release music because mixing is the bottleneck. Removing that wall — that's MUEDial's job.

The feel that mixing needs multiple rounds of back-and-forth — you only get that from being in the room. The reason I went with a chat UI is because the cycle of "text-based direction, revision, confirmation" maps perfectly onto how music direction actually works. I translated that real-world workflow directly into the product.