Efficient high-resolution image synthesis with linear diffusion transformer

(nvlabs.github.io)

220 points Vt71fcAqt7 | 1 comments | 16 Oct 24 14:56 UTC | HN request time: 0.201s | source

Show context

cube2222 ◴[16 Oct 24 17:58 UTC] No.41861846[source]▶

This looks like quite a huge breakthrough, unless I'm missing something?

~25x faster performance than Flux-dev, while offering comparable quality in benchmarks. And visually the examples (surely cherry-picked, but still) look great!

Especially since with GenAI the best way to get good results is to just generate a large amount of them and pick the best (imo). Performance like this will make that much easier/faster/cheaper.

Code is unfortunately "(Coming soon)" for now. Can't wait to play with it!

replies(4): >>41861942 #>>41863225 #>>41864501 #>>41865018 #

Archit3ch ◴[16 Oct 24 20:02 UTC] No.41863225[source]▶

>>41861846 #

If you generate 25x more images, you can afford to cherry-pick.

replies(2): >>41863739 #>>41864455 #

1. Lerc ◴[16 Oct 24 22:24 UTC] No.41864455[source]▶

>>41863225 #

That transfers computer time to user time. It's great when you want variations, less so when you want precision and consistency. Picking the best image tires the brain quite quickly, you have to take into account the at a glance quality without it overriding the detail quality.

I'd be curious to see how a vision model would go if it were finetuned to select the best image match to a given criteria.

It's possible that you could do O1 style training to build a final stage auto-cherrypicker.

↑