Fixed Training Configuration

Text Descriptions
Text 1:Cool and sophisticated, featuring crisp distorted guitar, shuffling drums and simple bass that create a vintage atmosphere.
Text 2:Powerful rock track with continuously improvised electric guitar solo. Played by a rock band of drums, bass guitar, electric guitar, acoustic guitar and electric organ. Energetic and lively - ideal for sports, action and motivational soulfulness. BPM 106
Text 3:Positive, tropical, shiny 4 tracks collection inspired by summer. Perfect for dance, energetic, party, travel or presentation video projects.
Text 4:An upbeat and fast-pace jazz sing trailer featuring horns, drums and vibes. This full-length and face-paced track tells a story.This song is perfect for trailers, promotional videos, advertisements, commercials, TV ads, YouTube videos, corporate presentations and wedding videos.

Text Description AR FM FM (VAE) GT
Text 1
Text 2
Text 3
Text 4

Controlability

The following examples are conditioned using temporally-aligned drums, chord progressions and melody controls that were extracted the ground-truth sample.

The figure paired with each audio corresponds to the extracted ground-truth melody condition.

Inpainting

Below there's a table containing multiple examples of inpainting, flip through the pages to see modre examples.