4 ms·
It has nothing to do with "style". DALL-E 2 consistently astonishes with its composition, framing, variety of visual elements, and so on, which tend to also be
by prvc 4y ago
It has nothing to do with "style". DALL-E 2 consistently astonishes with its composition, framing, variety of visual elements, and so on, which tend to also be good in oil paintings and professional photographs, but not at all in the sort of content which predominates on art sharing apps with user-generated content.
- orbital-decay 4y agoIf you mean porn (which is censored in DALL-E), SD also can't make it as it's mostly filtered out from the training set. All of it is made by custom finetuned models, which is my point. Same with hands, which SD can struggle with. Styles are not the only thing that you can transfer, but subjects and concepts as well.
- prvc 4y ago>If you mean porn I certainly do not.
- orbital-decay 4y agoWell... That's what predominates as user-generated content, I think :) Anyway, none of the existing image-gen models can be used in content production in their vanilla state. Prompt to image is fine as a hobby, but style/concept transfer is what makes them usable in professional setting. Or rather will make in nearest future, as this is all still highly experimental. SD in particular is quite small and is not a ready to use product, not intended for direct usage. It's a middleware model to build products upon. Such as Midjourney.