The model WAN2.2-ST-REAL-FEMBOYS-FIN2XX is versatile and was developed using a progressive improvement approach. Its distinctive feature is its hybrid architecture, resulting from the fusion of three different LoRA models. This is an initial test.
Description
FAQ
Comments (26)
"The presentation videos are created using the SMOOTH WORKFLOW v2.0 by Digital Pastel."
"The WAN2.2-ST-REAL-FEMBOYS model is usually used without any other LORA. It's up to you to decide if you’re not achieving satisfactory results."
Yes it is possible to make him appear his genitals even if your character and dressed in fact it depends on your invitation and your image to you to test there is what to do according to the examples of his videos I will test it and see what it gives
Do the test and use a lora as a supplement if it doesn't work. And don't hesitate to share your creations, thank you.
Please remember to publish the quantized GGUF versions for us low-vram boys. 🙏
Hi, I can't promise you anything because I don't have that skill.
Hello, Light. Your project is incredible; however, I’m afraid I don’t have enough knowledge to apply it properly.
Could you release the high/low version as you mentioned you would, along with a functional workflow using your creation? I tried using mine and the Smooth Workflow 2.0, but I wasn’t able to generate a decent video. It ended up very distorted and damaged the video. In addition, the penis coming out of the pants and not being visible in the initial image looked deformed.
I would like it to appear later and, if possible, not already be present in the initial image.
I’d appreciate it if you could give me a response. Thank you very much for the project.
Hi, the workflow is in the published videos. You just need to download the video and drag it into comfyui.
@lighthorsexajz830 I’m referring to a workflow. I don’t know if the videos let you download a workflow template exactly like that. But honestly, I might just be dumb and have never found this on Civitai.
Using both models WAN2.2-ST-REAL-FEMBOYS-FIN2XX and WAN2.2-ST-REAL-FEMBOYS-LOW-FIN2XX
I used your workflow with both WAN2.2-ST-REAL-FEMBOYS-FIN2XX and WAN2.2-ST-REAL-FEMBOYS-LOW-FIN2XX — I confirm they’re awesome. I didn’t think I could use them without any LoRA. Thanks for sharing.
Thanks for your interest and feedback — glad you liked it. It’s not perfect yet and still needs improvements, but I appreciate you sharing.
@lighthorsexajz830 Thanks
Absolutely wonderful. Works great!
Thank you for your comment, I really appreciate it and I'm very happy that you liked it. You are one of the few people who give their opinion. Feel free to share your creations on this post made with this model, thank you very much
Is this the high model?
Yes, that is correct.
Will this work with standalone Wan2gp or does it only work with comyui?
It works perfectly, you just need to put it in the right place
@lighthorsexajz830 thanks, i've really been struggling, is it simply detected in the right folder? I've read that some models require a json definition file for wan2gp to detect them properly. I'm really new to video models so im kinda clueless.
@CunningStunt I left wan2gp for comfyui, which is less complicated and easier to install
@lighthorsexajz830 no worries, I worked it out, the models go in the ckpts folder but you need to create a json with the same name and put it in the finetunes folder, then wan2gp is able to see the models.
@CunningStunt OK, great! Feel free to share your results on my post so our subscribers can enjoy them too.
@CunningStunt You'll need a JSON for WANGP
do note that this is specifically an I2V (it would be helpful if @lighthorsexajz830 updated their description/title to state that it's I2V, otherwise it doesn't work)
The JSON code is something like this:
{
"model": {
"name": "Wan 2.2 Futa Femboy I2V",
"architecture": "i2v_2_2",
"description": "Wan 2.2 Remix 14B FP8 quantized safetensor model with NSFW capability for image to video to make FEMBOY videos.",
"URLs": [
"ckpts/wan22STREALFEMBOYSLOW_v10.safetensors"
],
"URLs2": [
"ckpts/wan22STREALFEMBOYS_v10.safetensors"
],
"text_encoder_URLs": [
"ckpts/umt5-xxl/nsfw_wan_umt5-xxl_fp8_scaled.safetensors"
],
"auto_quantize": true
}
}
Note, I use a different text encoder. you can leave it out if you don't use a specific text encoder
Is there a workflow to get the most out of this? Sorry, still a bit of a noob, but this looks exactly what I was after,