Do you want to support me? Consider being a patron on Patreon to vote on train polls and access stuff early. ❤️
My first pose embed so it has some issues but I'll update it in future.
Known Issues:
1- Instead of cropping i added 512 512 white background on images and this causing white borders appearing on images time to time.
2?- Sometime Embed does too much gaping but it's up to your taste if it's an issue or not 0.0 ... You can try giving less weight to the embed like 0.8 for less gaping
Trained with 120 image and 3000 steps with clip skip 2. Tested on Anythingv3 and Corneo's 7th Heaven Mix
I hope you guys like it. I also open character polls and requests in my Discord server feel free to drop and say hi ^^
Description
1
FAQ
Comments (15)
How do you make these physical characteristic non-face T.Is? Would be interesting to make a rapunzel hair one, or wet hair
In characters you delete ones related to the character. Other stuff are same too, if you gonna train like here ass spread you delete everything tha related to the pose like ass, backview, ass focus etc etc when you start training AI will try to find similarities and make that a new keyword.
For rapunzel hair you can go with deleting "hair" based tags on training set. probably can work
"1- Instead of cropping i added 512 512 white background on images and this causing white borders appearing on images time to time."
You should be able to avoid that by tagging those images with something (like "white border" or "white letterbox"). They shouldn't show up as part of the TI keyword then.
You could also use the img2img tool to resize and fill rather than use the boders but tagging the borders should work just as well (or better).
You can try LoRA version much better https://civitai.com/models/13127/murkys-spread-ass-lora
Извините пожалуйста за тупой вопрос, куда это надо ставить?))
в (_._)
В строке Type указан тип, а рядом вопросительный знак. Там и написано
How to use it? What is the difference with murkys-spread-ass-lora?
btw both lora and textual inversion have "spreadassms" prompt naming, wouldn't they conflict with each other? As if you will use lora, textual inversion will also be loaded?
Yes if you have both in they will create extra gaping results. You can change TI name though
what version is it for?? sd 1.5 sdxl??? just says "other"
https://civitai.com/models/313623/spread-ass-embed This' you? It even uses the same sample picture you have here. He has some other stuff similar to yours too.
I don't see the image but trigger word ends with MS which was a signuture for early models. Thanks I'll report to Mods
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.