AI Mixing
AI mixing that shows its working
Give Euphona your stems and it returns a balanced starting mix — level, corrective equalisation and bus processing — with a reason attached to each move.
Meters one minute per minute — a four-minute track costs four minutes of your allowance.
How it works
What it measures, what it changes
And what stays yours to decide. Every stage is a parameter you can read back and overrule.
- 01
Every stem is analysed against every other
The interesting question in a mix isn't what each stem sounds like — it's where two of them are competing. So the analysis builds a masking map across frequency and time: where the kick and bass occupy the same space at the same moment, where the vocal is losing to the guitars. That map is what the moves are for.
- 02
Levels are set from the masking map
Static balance first, then automation where a fixed level can't hold the vocal above a dense section. The balance follows audibility, not a target level per instrument — 'turn it up' is what you do when you can't find the frequency that's in the way.
- 03
Corrective moves, then the bus
Per-stem corrective equalisation and dynamics to make room, then bus processing across the mix. Each move is a parameter you can see, change or discard, and each one is annotated with the masking it was addressing.
The specifics
What you would ask before trusting it with a record
Formats, limits, and the places it is honestly hardest. Nothing here is a range we have not measured.
- What this is
- An assisted first mix: a balanced, corrected starting point that a person then finishes. We deliberately don't sell it as a finished record. Full automatic mixing to release quality is unsolved, and claiming otherwise would be the fastest way to teach you not to believe the rest of this site.
- Inputs
- Your stems or multitrack, or stems produced by separation from a stereo mix. Metering is on the timeline duration of the mix, not the sum of the stem durations — twelve stems of a four-minute song is four minutes of processing, not forty-eight.
- The explanation
- Each move returns the masking measurement that prompted it. 'Cut 2.5 dB at 320 Hz on the rhythm guitar' is a setting; 'cut 2.5 dB at 320 Hz on the rhythm guitar, where it was masking the vocal for 60% of the chorus' is a reason, and a reason is something you can disagree with.
- What it will not do
- It doesn't make arrangement decisions, doesn't choose effects as production, and doesn't replace a mix engineer's judgement about what a song should feel like. Those parts of mixing aren't a measurement problem.
Questions
What people actually ask
- Is this going to mix my record for me?
- No. It gets you to a balanced, corrected starting point quickly and tells you what it did — the diagnostic part of mixing. What comes after that is taste, and we'd rather say so than pretend otherwise.
- Do I need separate stems?
- Ideally yes. If you only have a stereo bounce, separation can produce stems first, but a mix built from separated stems inherits whatever the separation left behind.
- How is this different from mastering?
- Mixing works on the parts and can change the relationship between them. Mastering works on the finished stereo file and can only change all of it at once. If the problem is one element, it's a mix problem.
- Can I take the moves into my own DAW?
- The report lists every parameter with the measurement that prompted it, so every move is readable and reproducible in your own session. The values are yours — nothing about the mix is locked to ours.
Try it
Put your own track through it
The free plan runs the whole engine on your whole song and plays you the result, matched for loudness against the original. No card. You pay when you want the file — one track at a time, or a plan if you are making a lot of them.
Who it’s for