For producers
Professional control, at the speed of an idea
The engine does the diagnostic pass in a minute and hands you every parameter it set. What you do with them is still your job.
What actually gets in the way
The reference bounce eats the session
A rough master so the artist can hear it in the car is twenty minutes you didn't budget, on every revision. Automating that specific twenty minutes is worth more than automating anything creative.
Automatic tools do not tell you what they did
A chain you can't inspect is one you can't correct, and it'll make the same wrong call on the next track for the same invisible reason. The output isn't the problem — the opacity is.
Every tool starts from zero
Separate in one app, tune in another, master in a third — and each one re-analyses the same audio from scratch, knowing nothing about what the last one changed. The work you put into teaching one tool about the track doesn't survive the export.
The workflow
How it goes
- 01
One analysis, reused
Loudness and dynamics over time, spectral balance, key and tempo, the transient map, a defect map and a genre embedding — computed once per track into a shared graph. Every capability reads that graph instead of running its own private pass. That's what makes chaining lossless.
- 02
Chain without re-uploading
Separate the mix, tune the extracted vocal, rebuild, master the result. The mastering stage reads what the mixing stage changed. There's no export-and-reimport step in the middle, and no point where the analysis gets thrown away and redone from a bounce.
- 03
Read the report, override the calls
Every job returns its analysis and every parameter the chain set: equalisation moves in dB by band, dynamics thresholds and ratios, limiter gain reduction over time, measured before-and-after loudness and true peak. Measured, not intended. Disagree with any of them and change it.
- 04
A/B level-matched, or do not bother
Comparison is always loudness-matched, because an unmatched A/B measures which one is louder and nothing else. That rule applies to our own before-and-after too — we're not going to win a comparison we rigged.
The suite
The parts you would use
Questions
What people ask before starting
- Does it work from stems or only from a bounce?
- Both, and the answer changes what is possible. From stems, mixing can move elements against each other. From a stereo bounce it cannot, because the elements are not separate any more — separation can produce stems first, and the result inherits whatever the separation left behind.
- Do I get the settings, or just the output?
- Every parameter the chain set, with the measurement that prompted it — EQ moves in dB by band, dynamics thresholds and ratios, limiter gain reduction over time. The values are reproducible in your own session; nothing about them is locked to ours.
- How is a long session metered?
- On the timeline duration of the material, not the sum of the track durations. Twelve stems of a four-minute song is four minutes of processing, not forty-eight. Analysis itself is never metered.
- Can I override what it chose?
- Yes, at every stage — the target, the reference, and each stage of the chain. The report exists so you can see a decision and disagree with it, which is not a useful feature if the decision is then fixed.
- Is there an API?
- Yes, async-only, with jobs, live progress and signed file URLs. If you are batching or wiring this into something else, that is the surface to use rather than the app.
Next