HOME | DD

Description
This is how Stable Diffusion renders Lucy Loud
Interestingly, Lucy is the single hardest character to pull off right in AI because she has no visible eyes. AI programs don't like that at all.

A preliminary test to see how well AI can handle the Loud House. Even when trained on 256 images in DreamBooth, the best Stable Diffusion can do is isn't that great unless you add loads of qualifiers. Craiyon is, beyond all belief, the best at replicating the Loud House's style, but it's so pisspoor in terms of quality that you have to upscale it with another program, but no other program understands the Loud House and invariably screws it up when doing so.

DALL-E is the absolute worst. Utter dogshit. Great at some things, but awful at the Loud House. If you could train DALL-E 2 on Loud House images, it might work better, but so far that's just not possible.

Someone ought to train a Stable Diffusion model purely on the Loud House. Then there might be some progress.

Description
This is how Stable Diffusion renders Lucy Loud
Interestingly, Lucy is the single hardest character to pull off right in AI because she has no visible eyes. AI programs don't like that at all.

A preliminary test to see how well AI can handle the Loud House. Even when trained on 256 images in DreamBooth, the best Stable Diffusion can do is isn't that great unless you add loads of qualifiers. Craiyon is, beyond all belief, the best at replicating the Loud House's style, but it's so pisspoor in terms of quality that you have to upscale it with another program, but no other program understands the Loud House and invariably screws it up when doing so.

DALL-E is the absolute worst. Utter dogshit. Great at some things, but awful at the Loud House. If you could train DALL-E 2 on Loud House images, it might work better, but so far that's just not possible.

Someone ought to train a Stable Diffusion model purely on the Loud House. Then there might be some progress.