I wonder if you could use another model that describes music and feed that text into this one?
Even something based on spotify's music labeling api would be super interesting!
I will get excited when I see this making images bigger than 256 pixels.
Currently both Big Sleep and Deep Doze are generating 512x512. These ones are representative: https://postimg.cc/HVspWgPn
Few are "collages", most images have full-area coherence.
> as album cover art
Indeed, what I see is an album cover generator.
“A man painting a completely red image” is very much a dadaist collage. The only complaint is that the ‘man’ could be rather more recognizable as such.
Is there a way to perform a similar translation with music? For example, if you play in D Minor (the saddest of all keys), is there a way to map the key or some other musical characteristic to a word and have the images be generated with the intermediate being the primary source? Or would the approach be to map images to certain characteristics of music directly?