ShopDreamUp AI ArtDreamUp
Description
Looks like this blob of a beast requires another top-up. Care to help them indulge?
The conversation in Furry Diffusion turned towards the topic of blob generations, and I was inspired to see how large I could generate a character without any manual guidance. Turns out, pretty dang blobby! The main way to get larger sizes (which I've yet to include mention of in my fats guide >_>) is to exploit the AI's urges around character framing. It's trained to fit a character within the bounds of the picture, at distances comfortable for viewing at, but this has the side effect of the aspect ratio (largely width) from limiting how fat a character can be (their body can't stretch out of bounds). Using a wider aspect ratio therefore helps prevent the weight being clamped, with LoRAs making the AI comfortable with more zoomed out characters also helping to form blobs.
The one issue I was having with this is tails. The AI liked to dedicate significant portions to the dragon's tails - in some cases, half the image was chonky tail. Unfortunately even negativing tail, and positiving tailless, didn't seem to help much. Probably a limitation of the model :v
Details listed are for the original image, not this touched up version.
Prompt: full-length portrait, front view, solo, anthro, dragon, kemono, female, green scales, morbidly obese, immobile, (ussbbw:1.4), <lorasymodelBiggerGirls_v30NewTagging)-(NovelAI):1.5>, fat rolls, medium hair, messy hair, black hair, featureless breasts, sitting, hyper belly, hyper butt, (tailless:1.2), BREAK by flipside, by chibitay
Negative prompt: cub, bloated, (tail:1.2)
Other: Steps: 25, Sampler: DPM++ SDE Karras, Cfg scale: 7, Seed: 493788942, Size: 2376x768, Mode: Txt2img, Clip skip: 1, Hires factor: 1.5, Hires strength: 0.6, Hires upscaler: Lanczos, Cfg rescale: 0.75, Prediction type: V
Model: EasyFluff V11.2 [821628644E]
The conversation in Furry Diffusion turned towards the topic of blob generations, and I was inspired to see how large I could generate a character without any manual guidance. Turns out, pretty dang blobby! The main way to get larger sizes (which I've yet to include mention of in my fats guide >_>) is to exploit the AI's urges around character framing. It's trained to fit a character within the bounds of the picture, at distances comfortable for viewing at, but this has the side effect of the aspect ratio (largely width) from limiting how fat a character can be (their body can't stretch out of bounds). Using a wider aspect ratio therefore helps prevent the weight being clamped, with LoRAs making the AI comfortable with more zoomed out characters also helping to form blobs.
The one issue I was having with this is tails. The AI liked to dedicate significant portions to the dragon's tails - in some cases, half the image was chonky tail. Unfortunately even negativing tail, and positiving tailless, didn't seem to help much. Probably a limitation of the model :v
Details listed are for the original image, not this touched up version.
Prompt: full-length portrait, front view, solo, anthro, dragon, kemono, female, green scales, morbidly obese, immobile, (ussbbw:1.4), <lorasymodelBiggerGirls_v30NewTagging)-(NovelAI):1.5>, fat rolls, medium hair, messy hair, black hair, featureless breasts, sitting, hyper belly, hyper butt, (tailless:1.2), BREAK by flipside, by chibitay
Negative prompt: cub, bloated, (tail:1.2)
Other: Steps: 25, Sampler: DPM++ SDE Karras, Cfg scale: 7, Seed: 493788942, Size: 2376x768, Mode: Txt2img, Clip skip: 1, Hires factor: 1.5, Hires strength: 0.6, Hires upscaler: Lanczos, Cfg rescale: 0.75, Prediction type: V
Model: EasyFluff V11.2 [821628644E]
Image size
2376x768px 2.25 MB
Created using AI tools
Comments2
Join the community to add your comment. Already a deviant? Log In
Lovely!~