3 ms·

I have a penchant for wanting to make technically "bad" or heavily stylized photos - and Stable Diffusion is pretty poor at those. There's very little good boke

I have a penchant for wanting to make technically "bad" or heavily stylized photos - and Stable Diffusion is pretty poor at those. There's very little good bokeh or tilt shift stuff and CCTV/Trailcam doesn't come out too well. In fact Dall-E isn't as impressive for some styles as "older" models (Jax/Latent Diffusion etc)

vhold 4y ago
My hunch is that is the result of this: https://github.com/CompVis/stable-diffusion#weights https://github.com/CompVis/stable-diffusion#weights > 515k steps at resolution 512x512 on "laion-improved-aesthetics" (a subset of laion2B-en, filtered to images with an original size >= 512x512, estimated aesthetics score > 5.0 https://github.com/LAION-AI/laion-datasets/blob/main/laion-aesthetic.md https://github.com/LAION-AI/laion-datasets/blob/main/laion-a... for more details. What's remarkable is this: https://github.com/LAION-AI/laion-datasets/blob/main/laion-aesthetic.md#training-set https://github.com/LAION-AI/laion-datasets/blob/main/laion-a... That aesthetic predictor was apparently trained on only 4000 images. If my thinking is correct, imagine the impact those 4000 ratings have had on all of the output of this model. You can see samples (some NSFW) of different images from the original training set in different rating buckets here, to get an idea of what was included or not in those training steps. http://3080.rom1504.fr/aesthetic/aesthetic_viz.html http://3080.rom1504.fr/aesthetic/aesthetic_viz.html