iamddtla ah, ok.. just curious, what models did you use for the merge?

V6 was a big upgrade from previous versions. V7 seems to have better prompt adherence and can handle a wider range of concepts while still maintaining good quality. It can handle some PONY loras as well without looking over-exposed. Like any checkpoint some prompt combinations give more issues with anatomy. I've also gotten results that I've NEVER seen before short of Big Asp but with much better quality. It's pretty good about maintaining a stable character; if you want more randomness try the reddit prompts. I've gotten some stunningly good results that are better than any checkpoint I've ever used for realism and artistic quality - kind of like 'Dang, that's the most beautiful image I've ever seen of a pretty girl getting jizz-blasted by a gang of construction workers!' - lol. Does just fine with softer core stuff too. I haven't tried any lesbian prompts yet; everything falls short on that so far. It didn't do pierced nipples either. Pubic hair works well, It DID accurately create tattoos of a specific word about 50% of the time, which is pretty astounding. Some PONY tattoo loras work too without getting washed-out images. Freckles look good. I would call V7 the current state of the art for NSFW, honestly.

V7 simply has way better results. It's day and night.

only if you have a 4090 or above. otherwise just generate pics. unlike to generate pictures, "barely run it" doesnt work on generating videos, you would get terrible results.

XpPillow This is video is generated by kling, no graphic card needed

using https://civitai.com/models/1824594/rapid-wan-22-all-in-one-updated-v8?modelVersionId=2064786 I can generate an 81 frame 480p video in 2 1/2 minutes on a 4070. just need to offload.

@prekillo484 what ?
kling ai online ? or local offline
it doesnt support nsfw ?

well these are text to image models. at least I don't think they have an t2v model on this page?

dow185 But how are the videos that are visibly using this model being made?

Kyrandar with a separate image-to-video model. it's become a trend to animate the preview images for text to image checkpoints and loras. I use this model, it works great on my 5070: https://civitai.com/models/1824594?modelVersionId=2064786

there is a illustrious version of this model in early access, available soon.

@iamddtla not on the same level of detail and realism yet

@iamddtla Oh ok? I mean the first amateur model is pretty distinct from the pro So from my having using them, the pro models seem quite different? Or did you just merge amateur stuff into those models? Can you explain what you said in terms of specifically how it relates to the content in the models?

Thanks

@beepbopbip araminta fv6 have been great for amateur candid looking stuff.

@iamddtla Thanks for all your incredible work, I am bit confused at this point as you work so hard and now there are multiple millions of checkpoints<>loras! - I

So in your description - Can you clarify which model or Lora you was refering to that plays best with age ranges? - Thanks for your time!