9 ms·

My best (frankly uneducated) guess would be that they trained the model with famous movie posters as the objective, and still frames of the movie as the trainin

by a1371 5y ago

My best (frankly uneducated) guess would be that they trained the model with famous movie posters as the objective, and still frames of the movie as the training data. Then they gave frames of these movies to the ai to get a poster out.

suction 5y ago
Oh, so image data? I got the impression they used a text-only description as input.
yreg 5y ago
I don't think they look as if movie posters were the training objective. No texts, no large faces of leading actors, wrong aspect ratio, unusual colour palette.