AI mastering without the black box
Automated mastering analyses your file, compares it against targets learned from released music, and applies EQ, compression and limiting. The processing is usually competent. The frustration people describe is almost never about the processing.
The thing an algorithm cannot know
A mastering engineer asks a question first: what is this for, and what bothers you about it? You say the low end feels loose, or you want it harder than the reference, or the hats are fatiguing. That single exchange does most of the work.
An automated service gets none of that. It has your audio and nothing else, so it has to guess your intent from the file. Sometimes it guesses right. When it guesses wrong, you get a result that is technically fine and simply not what you meant.
That is the black box problem. Not bad processing — no conversation.
Why one-shot rendering makes it worse
If every adjustment costs a render and a wait, you stop adjusting. Two or three tries in, most people accept whatever they have and move on.
Mastering is not a decision, it is a series of small comparisons. Slightly brighter or not. Slightly harder or not. Each one takes seconds to judge and only makes sense in immediate contrast with the alternative. Break that loop with a progress bar and you have not slowed the process down — you have replaced it with something else.
How to judge a master, whoever made it
Whether you use a service, a plugin chain or an engineer, the checks are the same:
- Match the loudness before you compare. Non-negotiable. Louder wins every blind test for the first ten seconds, and ten seconds is all it takes to fool yourself.
- Low end: does the kick still have a front edge, or has it turned into a round thump? Compression takes the attack first.
- Top end: bright is not the same as detailed. Listen for whether hats and air sound clearer, or just louder and harsher.
- Depth: in an over-limited master everything sits at the same distance. If the reverb tails have flattened into the dry signal, back off.
- The main element: does the lead or vocal still sit where you put it in the mix, or has the master moved it?
- Check it small. Phone speaker, laptop, one earbud. Problems that are subtle on monitors are obvious there.
What no mastering will fix
Mastering works on a finished stereo mix, which means it moves broad strokes. It can shape balance, control dynamics and set level. It cannot separate elements that are fighting each other, fix a muddy low end caused by two instruments occupying the same space, or repair a harsh vocal.
If a master keeps sounding wrong no matter what you do to it, the answer is usually in the mix. That is not a limitation of automation — a human engineer would tell you the same thing, just more expensively.
What the alternative looks like
We built ZERRRO Engine around the feedback loop rather than the render. You pick a genre and a character, move the dots, and hear the result as you move them — no queue between one idea and the next.
When it sounds right, you compare it against your original mix at matched loudness, and only then does anything get made loud.
Hear the difference an adjustment makes
No file on your phone? There are demo tracks in the tool — drag the character and listen to what changes.
Try it →
0 comments