The goal of this model is to create slightly more "realistic" female growth sequences. This means that the size and body change is less extreme. The model can do "full body", "upper body" and "cowboy shots", but the quality drops the further away the subject is. Theoretically "fmgs" (female muscle growth sequence) should correspond to transformation, with the subject having the same clothes and background, while "fmgp" (female muscle growth progress), should correspond to progress pictures of a transformation lasting a longer period of time, with different clothes, poses and background. In reality either prompt can do both.
There are other transformation models out there that can be combined easily and also muscle models, to increase the size of the subject.
V1 was trained on Almost Real HotDog Mix (Pony) using ~100 FMG Sequences, including a lot of realistic ones.
V2 was trained on WAI-Illustrious using ~300 realistic FMG Sequences. (also optional version of Hassaku).
Anima V1 - Trained on cartoon / anime FMG images
Krea 2 - V1 - Trained on a mix of realistic and cartoon images. Only FMGS and "a female muscle growth sequence" were used as tags for training.
Krea 2 - DESC - Same as V1, but used automated natural language description.
Krea 2 - TAG - Same as V1, but used automated tags.
Recommended prompts to start (SD Versions): "FMGS, transformation sequence, transformation,muscle growth,1girl,score_9, score_8_up, score_7_up, score_6_up,realistic, upper body, black hair, facing viewer"
OR
"FMGS, transformation sequence, transformation,muscle growth,1girl,score_9, score_8_up, score_7_up, score_6_up,realistic, full body, blonde , facing viewer"
Recommended negative prompt: "ugly face, score_4,score_3,score_2,score_1,worst quality,bad hands,bad feet,"
Recommended prompts to start (Anima, Krea 2 Versions): See examlples, long detailed prompts work best, but add keywords like "FMGS" and "A female muscle growth sequence" to better trigger the LoRa
Recommended Weigth: 0.3 - 0.85
Recommended dimensions: 1024 x 584 (best for full body), 1024 x 432 (best for upper body), 1344 x 768 (best overall), 1216 x 832, 1536 x 640 (best for multiple stages), 1024 x 1024 (worse)
Recommended LoRas to combine (SD): Gravs Transformation & Muscle Mommy
Recommended LoRas to combine (Krea 2): Wellness IFBB , Muscle Slider, Muscular Women , Transformation
See the example images for inspiration (up scaling, can fix issues with faces, bad quality etc)
Description
Same dataset as V2, but trained on Hassaku XL (Ilustrious)
FAQ
Comments (2)
i like the concept but the faces are always about the same no matter what I prompt or LORAs I use. seems too baked for both the faces and expressions.
Thank you for the feedbacK! Yes, there are issues with the faces, but not sure how to fix them for a future version. I have noticed that closer shots (e.g. upper body only) and cartoon instead of realistic do perform better, so that is an option. You can see some of the posts here for prompts and use them as a base for what you want to generate. Hope this helps.