@Civitaiwolfxx means"fxxx!" of course it is a good sense

Note: I bet the "extreme close-up" training word would go great with the hard cut lora 👀

Bro, TY

For me adding in the oral insertion lora at 0.5 works well.

https://civitai.com/models/1874153/oral-insertion-wan-22

If you use the workflow he posted in one of the other models/versions, you should get similar results. I think the split was 3/3 high low

@Civitaiwolfxx Thanks!!

With a combination of smoothmixwan checkpoint and this lora, I was able to finally get him to hold it in deep for the entire 5 seconds. Some of the OP's prompts regarding holding her head in place and stuff work for me.

@HNKenshiro thanks. going to give that a try

You know what I've found? When I do a continuation deep throat hold video with the last frame of a previous video that was also a deep throat hold, and when stitch them together, it almost looks as though she's trying to pull out and he forces her right back in. Looks very rough.

I want to know the image prompt

Just an idea: Since it´s possible to change the cum with trigger words (the cum is sticky, the cum is thick and sticky, the cum is translucent, the cum is watery), if you generate multiple clips you might get away with triggering the tears as translucent or watery and in a later clip the cum as.. cum. Download the training data if you haven´t already, it´s really helpful with this lora.

I got it working

HIGH
nsfw/wan_2.2_12y_highnoise_broken_v1.0.safetensors 0.60

Wan21_12V_14B_lightx2v_cfg_step_distill_lora_rank64.safetensors 3.00

nsfw/Wan22_ThroatV3_High.safetensors 0.55

LOW
nsfw/wan_2.2_i2v_lownoise_broken_v1.0.safetensors 1.00

Wan21_I2V_14B_lightx2v_cfg_step_distill_lora_rank64.safetensors 1.50

nsfw/Wan22_ThroatV3_Low.safetensors 0.40

PROMPT:
say "she is crying with clear transparent teardrops"

Negative:
oversaturated colors, overexposed, static pose, static image, still image, motionless, blurry, low detail, unclear details, subtitles, text, watermark, signature, in the style of, artwork, painting, illustration, grayscale, desaturated, worst quality, low quality, bad quality, jpeg artifacts, compression artifacts, ugly, deformed, disfigured, mutilated, bad anatomy, missing limbs, extra limbs, extra fingers, fused fingers, mutated hands, poorly drawn hands, poorly drawn face, malformed limbs, three legs, crowded background, cluttered background, messy background, inverted, upside down, blurry tears, opaque tears, no tears, absent tears, indistinct tears, fake tears, thick tears, viscous tears, stringy tears, milky tears, white tears, cloudy tears, murky tears, gel-like tears, sticky tears, semen-like tears, cum-like tears, tears resembling semen, tears resembling cum, tears like ejaculate, white fluid tears, thick fluid tears from eyes, viscous fluid from eyes, semen tears, cum tears, milky tears, sticky tears, stringy tears, gooey tears, thick liquid from eyes, white discharge from eyes, semen from eyes, cum-like fluid from eyes, ejaculate-like tears, opaque eye fluid, viscous eye discharge, gelatinous tears, slimy tears, phlegm-like tears, mucus-like tears, blood, bloody penis,

@lifter969 Thanks this is a great idea

Most people are unaware but Neg prompts are broken in virtually all Wan 2.2 workflows as people are using SD Wan 2.1 workflows that were converted to Wan 2.2 - you can remove all Negs and it unfortunately may not have any impact on your generations - just a PSA

Got past my issue :D, though all the videos, the woman gives a blowjob but never deep. I have tried a good number of prompts but no luck.

@Civitaiwolfxx Yes with your workflow (btw, love the multi-prompt setting, really useful) getting some good results, just not the deep aspect oddely, unless I missed a step with the install. I followed the instructions (though the default was 8Q not 4Q as the instructions said, but got the 4Q). I will try again. I used all your prompts (again thank you, VERY VERY helpful, and also thank you for all your work with the workflow).

Also forgot to add, is there a way to add neg prompts to the workflow?

Also, if I was to try and use a Checkpoint, how would I load it into this workflow, getting some good results at the moment though, so thank you and the workflow is so easy to read and use, again thanks :D

for video typically you need a longer one. You can even tell it how many seconds for each portion etc. Sometimes you can just leave it alone and the lora might do the work. In my experience the best ones have a lot of text. Some of it is in another language.

use comfyui or forge/automatic1111 I think forge is waaaaay easier.

give me one :D

@Civitaiwolfxx I cannot seem to find it (nothing with Chinese characters within "Cumworkflow" that is linked) , I can see the 4 pos prompts on the left hand side, followed the K sampler to a node called prompt and moved everything around but no luck :D. Either way, still doing an amazing job, just get odd blurryness at times.

Nevermind, my bad, I was using your older version, I will try the newer one :)