Character LoRAs - Want more character variety? Then check This List!
Description

Experimenting with Wan image to video locally, generated on my video card. (Before I used the wan nsfw lora, there were assorted issues with genitalia…)
Both the starting image and the video were generated by me, so I’m giving both prompts.
The image prompt was:
score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, source_pony, rating_explicit, cute furry feral pony Babs Seed giving a human male a blow job, side view, drool, human on pony action, hand on head, grab, throat bulge, penis, curtains, bedroom, bed, head focus
Negative prompt: signature, watermark, monochrome
Steps: 30, Sampler: gradient_estimation, Scheduler type: sgm_uniform, CFG scale: 5.5, Seed: 4, Size: 1024x1024
This then was sent to Wan 2.1 14b 480p i2v, Q4_0 GGUF version, 512x512, length 33, with the lora “wan-nsfw-e14-fixed”, Model Sampling SD3 7.0, seed 3, steps 20, cfg 5.5, uni_pc simple, vae tiled set to 256/64/64/8. That got fed through an upscale image node to bring it back to 1024x1024. And I had to convert the webp to a gif, because the site didn’t like webp.
Positive: Pony sucking a human’s dick, eyes rolling back. The human has his hand on the pony’s head, pulling her head towards him.
nsfwsks
Negative was the default in the example workflow, and wasn’t in english, so hopefully the characters appear correctly: 色调艳丽,过曝,静态,细节模糊不清,字幕,风格,作品,画作,画面,静止,整体发灰,最差质量,低质量,JPEG压缩残留,丑陋的,残缺的,多余的手指,画得不好的手部,画得不好的脸部,畸形的,毁容的,形态畸形的肢体,手指融合,静止不动的画面,杂乱的背景,三条腿,背景人很多,倒着走

Tags
explicit19483 ai generated41368 generator:prefect pony v5 xl27 generator:stable diffusion21356 generator:wan v2.13 prompter:lord waite65 babs seed48 human8420 pony20385 g435049 animated300 blowjob1263 blushing10628 curtains503 drool957 drool string243 female38930 foalcon954 hand on hair5 hand on head181 human on pony action2016 interspecies2878 male8863 nudity21494 oral1888 penis6503 prompt in description1019 sex6155 side view342 straight6658
Source

Comments

Syntax quick reference: **bold** *italic* ||hide text|| `code` __underline__ ~~strike~~ ^sup^ ~sub~

Detailed syntax guide

@derp621
Not all that exotic. I’ve got a 3060 12GB with 32GB of RAM.
As far as time, it looks like 12-15 minutes, bearing in mind that it didn’t have to run the initial image generation for most of the runs. I probably would have had more trouble if I wasn’t running a GGUF version of the model, and I kept the video generation at 512x512 and upscaled because I didn’t think it would have worked for larger resolutions. I probably could have run it through a refiner of some sort, since anything that upscales an image would have worked there.
I’ve got a fair number of bad results from before I tried that lora, most of which were almost looking alright.
…and so it begins… local video/gif generation.
How long did it take? Do you need exotic hardware or can it fit on a consumer GPU?