It hears the problem
Every move starts from a measurement of your recording, not a template. A noisy take gets a gate set to its own measured floor. A clean one is not gated at all.
VoxBrain 1.0 · VST3 & Standalone
One click.
One click.
Done.
VoxBrain listens to the vocal in front of it, measures what is actually there, and builds the whole chain around those numbers — tone, dynamics, colour, space. Then it tells you exactly why, in words you can argue with.
Perpetual licenceFree updates
No subscriptionRuns offline
What it actually does
Not a preset.
01
02
03
Not a preset.
An engineer that listened.
Most “AI mixing” is a preset with a marketing budget — the same chain whatever you feed it. VoxBrain measures your take first: level, noise floor, tone balance across seven bands, sibilance, dynamic range, how busy the transients are. Then it decides, and shows its working.
It explains itself
The report opens with what I heard before what I did. You get the evidence, not just a conclusion — so you can disagree with it, and learn from it.
It leaves your work alone
Lock any module and the AI will not touch it — not on auto-mix, not from chat, not from a preset. And it tells you what it left alone.
The workflow
Three clicks
Three clicks
to a finished vocal.
A verse is plenty. Press LEARN, play it, press it again.
- Pick a genre first if you want — 46 of them, from Drill to Neo Soul to Broadcast. Leave it on Auto and it works the style out itself.
- The chain is built and playing before you have finished reading the report.
- One Undo reverts the whole pass. Not one parameter at a time.
What I heard
Low-mids sit 6 dB above your core tone — that is the mud.
Sibilance ratio 0.71, well above comfortable.
Noise floor at −34 dB: the room is in the take.
What I did
Cut 3.2 dB at 310 Hz, narrow, to clear the mud without thinning the body.
De-esser at 7.4 kHz, keyed to your own sibilance.
Gate at −28 dB — 6 dB above your measured floor, so breaths survive.
An illustration of the report's shape. Your numbers will be your own.
Ask it anything
Type “make it brighter” and a knob moves instantly. Ask “why does my vocal sound muddy?” and it answers from your own measurements, quoting your numbers rather than generic advice.
- 89 production concepts, 46 genres and every module, shipped in the binary
- Follow-ups work — ask “why?” and it knows what you meant
- Answers the part you asked about, not five paragraphs
- No account, no subscription, no network round trip
The assistant
It runs
It runs
offline.
Everything it knows ships inside the plugin. Connect your own OpenAI, Anthropic or Gemini key if you would rather a large language model handled the conversation — but commands stay local and instant either way.
Formats
VST3 · Standalone application
System
Windows 10 or 11, 64-bit. A CPU from 2014 or later.
Licence
Perpetual, two machines, free updates.
macOS is in development. If you want it, say so — that is how the order gets decided.
Hear it on your own voice.
Sixty dollars, once. Perpetual licence, free updates, and fourteen days to change your mind for any reason at all.