02 — 2026
MixMentor
A web app that listens to a mix the way an engineer would — extracting real acoustic features and turning them into genre-aware, timestamped feedback instead of generic notes.
The problem
Most AI mixing feedback is generic because it never actually looks at the audio. MixMentor grounds every note in measured signal data first.
What it does
01Extracts acoustic features — spectral centroid, RMS energy, dynamic range, frequency-band distribution — with Librosa, and feeds the structured signal data into GPT-4 for genre-aware mixing feedback.
02A prompt schema converts raw signal analysis into timestamped feedback cards, grounding model responses in quantified audio characteristics rather than generic output.
03A reference-track comparison mode: upload a professional mix alongside your own and get a delta analysis of the key acoustic differences.