Generation Settings
Protein input is reverse-translated with the selected species codon table; DNA/RNA retains its codon-derived rhythm levels.
No file selected
If no PDB is uploaded, ESMFold predicts a structure. If no sequence is entered, the sequence is extracted from the uploaded PDB.
Species-specific codon preference affects reverse translation and the four-level rhythmic mapping.
World modes retain source root, third, fifth and seventh roles, then realize them as profile-appropriate rolls, arpeggios, drones or layered sonorities.
4.0 · Pure Physics
0 · JazzGPT4 · Pure Physics (Model A)
The model changes automatically. Values below 4 blend JazzGPT with physical scoring; 4 uses pure Model A.
Audio tools: fluidsynth · HiDef.sf2 · ffmpeg
How It Works
1

Read the biological input

Use DNA/RNA, a protein sequence, a PDB structure, or a sequence–structure pair.

2

Resolve spatial neighbours

The uploaded PDB is analysed directly; without one, ESMFold predicts a structure.

3

Map biology to music

Codons control four rhythm levels while amino-acid properties and 3D neighbours select source chords whose root, third, fifth and seventh remain traceable through modal projection.

4

Listen and refine

Play the generated audio, inspect the 3D structure, then adjust timbre and volume in the mixer.

Protein Structure
The uploaded or predicted protein structure will appear here.
Generating Music
Submitting the job…0%
Generated Music

Track Mixer

Choose timbres for Melody, Bass and Chord, adjust every available role, then create a separate MIDI and MP3 render. Percussion retains its profile-specific channel-10 pattern.
Master EQ

Vote for a Version

Select the original or a newly rendered mix. The original is never overwritten.