Update March 2026:
I recommend my new LoRA Wan 2.2 Close-Up Facial Cumshots over this one.
The purpose of this LoRA is to produce cum that appears thicker and slimier. I wanted to create this because I found that aspect lacking in other resources. It is trained almost exclusively on tight-framed facial imagery so performance is likely not good in other usages.
The style of cum and lighting in the input image seems to influence the slimyness/thickness generated in the output. These attributes seem to come into play more when the input image has them as well. In other words it performs better when there is thick slimy cum in the start image.
Combine it with Wan Cumshot for better control over shooting; however this degrades the performance in my experience. Again an input image that triggers the inteded properties in a solid way seems to help a lot. I haven't combined it with other LoRAs than Wan Cumshot yet so maybe there are better combos.
Version 3 has evolved from two prior unreleased versions. V3 is the first release.
It is trained on the following phrases:
cum shoots from a penis
cum on {face|tongue}
cum in mouth
{static|handheld} cameraExample prompt:
<other text describing scene, actions etc> cum shoots from a penis. cum on face. cum on tongue. cum in mouth. static camera.I'm looking forward to see what others will create!
Description
It has evolved from two previous unreleased versions. This is the first release.
It is trained on the following phrases:
cum shoots from a penis
cum on {face|tongue}
cum in mouth
{static|handheld} cameraExample prompt:
<other text describing scene, actions etc> cum shoots from a penis. cum on face. cum on tongue. cum in mouth. static camera.FAQ
Comments (60)
You can refer to my work, but most of it is animation, so it can have a dense or rich feel. I think prompt words are very important.
Here are the three LORAs I frequently use, along with their strengths.
These three are always used together.
strength:
Wan Cumshot : 1
Facials - Wan2.1 T2V 14B : 0.7
Wan 2.1 Facials - Cum on face, Bukkake, Lora, T2V, I2V : 0.5
Thanks I will try. It does look thick in your generations. I've tried them separately before but not enough to find a good mix.
I've not learned how LoRA weights stack. I tried your stack plus mine with my usual 1.5 as weight and that turned out real bad quality. But hinting at cool properties.
My second attempt I used
Thiccum: 1.0
Wan cumshot: 0.5
Facials: 0.5
Cum on face: 0.5
And that turned out good image quality-wise. I'm not sure how much of Thiccum contributed or changed what would have been otherwise because I don't know the others too well. But definitely interesting.
@lethotrisv439 Regarding "wan cumshot": From my perspective, this term provides the trajectory of the shot. However, the downside is it can sometimes result in a missed shot, unless you start very close to the face. Personally, I prefer a shot from a bit further away.
For "facials": This clearly specifies the target location: the face.
As for "cum on face": My personal experience with this term is that it conveys a sense of large quantity and thickness.
I haven't tried your LORA yet, but I'll give it a go when I have time to see the difference. Mainly, my current approach already consistently achieves the results I'm looking for.
I rarely use real people in my creations, though.
@lethotrisv439
For real people, I'd suggest sticking to these two for the best results. You can also add Detailz-Wan: Detail Enhancer for WAN Videos.
Here are the settings I'd recommend trying:
Wan Cumshot: 1
Facials - Wan2.1 T2V 14B: 0.7
Detailz-Wan: Detail Enhancer for WAN Videos: 1
maybe add penis lora , you can choice
@FLOW0308 Thanks a bunch that will save me a lot of time.
@lethotrisv439 I'm actually trying out your LORA now! I just thought of a good way to test it. If it works well, I'll upload the results so you can see how it turned out. I'm running it while watching anime, haha.
@FLOW0308 Nice! Can't wait to see what you can do. But for now I'm off to sleep.
Appreciate you sharing! These kinds of posts are very helpful to figuring out the perfect mix of loras
@lethotrisv439 I just tried your LORA. At a setting of 1, the effect was too strong and didn't look natural to me, so I adjusted it to 0.5, which is much better.
However, the quantity might still be a bit too much, giving it a very "silky" feeling on the face. I've uploaded some examples, and it looks pretty good with the anime style. I'll probably need to do more testing.
Still, for anyone who likes a lot of volume, this LORA is really useful! You'll just need to experiment with the weights yourself.
@FLOW0308 My recommendation of strength 1.2 - 1.8 might be too high when combined with multiple other LoRAs... but of course it's down to personal preference too. But I find that image quality worsens, sometimes drastically, and unwanted objects etc appear when the sum total of LoRA weights gets too high.
I don't know how this works. A sum total that works with one prompt and input image doesn't work with another. I'm guessing because the LoRAs involved activate to different degrees and at some point come into conflict. I probably need to learn more about the internals to understand this.
About the quantity, yes it's trained on a dataset with some large quantities. Another factor could be the framing. It is trained on quite narrow angles, so if applied on wider frame inputs I think it may possibly just scale up.
@lethotrisv439 For me, a total weight of 2.2 is the limit, but a total weight of 2.0 is the perfect balance. Of course, I'm just using the simplest addition to calculate it.
I prefer to have one main subject with the highest weight, usually around 1 or 0.7. The second highest weight depends on how I want to define my prompt words. The last one is for auxiliary support. I'll only use a maximum of four LORAs; any more than that, and I find it easily breaks down.
Your LORA can replace two of mine because your LORA has a static camera setting, which makes it excellent for auxiliary support. Sometimes Facials - Wan2.1 T2V 14B tends to move the camera. Wan 2.1 Facials - Cum on face allows me to express a thick, heavy feel, and its viscous texture looks great on skin, which I really like. So, I always use it as my final auxiliary.
Now, your LORA can be used to perfectly position the camera and achieve that thick, viscous feel. All I need to figure out now is how to make it even more viscous. Do you have any pre-set prompt words related to viscosity?
@FLOW0308 Funny that you mention viscosity. I avoided using it in the description to keep it simple but I use it all the time in my own documents and whatnot. There were no specific tags used for viscosity in this. I opted for very simple prompting in v3 but in v1, my first training, I used gelatinous cumshots and gelatinous cum. I've dropped that because all/most of the dataset has this property. I reason it should be applied by default by my LoRA, at least that's what I want. But this choice must also mean that it gets into tug-of-war with other LoRAs that associate with cum; for better or worse. But I think it's the right choice.
It's great to hear that my LoRA can replace two in your workflow. As for the overall thickness — that's my main focus here so I hope to improve it further in coming versions. For the next I'm trying to emphasize it even more in the dataset. I'm considering tagging opaque cum vs. hazy/translucent cum though, because both of those are present and I want control over it.
I trained four versions of this LoRA over quite a short span of days so I haven't had time to evaluate them in full depth. V3 looks most promising so far but I'm going to test the others a bit more too.
Unloading such heavy artillery at point blank range must surely be a war crime.
a noob question , does wan i2v maintain the original face of input image ?!
In my experience yes the base model does that quite well.
How to use the suggestions regarding the Wan Cumshot series!! | Civitai
I've written a reference article that everyone can check out, and you can also refer to it in conjunction with my work.
Can anyone get this working with Causvid, it seems to remove a lot of fluid dynamics
Try two Ksamplers. 1st one without the Causvid lora for a few steps, then the 2nd with the Causvid lora.
Refer to this workflow https://civitai.com/models/1622023/causvid-2-sampler-workflow-for-wan-480p720p-i2v .
Cheers ill give it a go
@hishiryo Thanks i'll give it a go
@Trollmale101 did it work? im using the same workflow
@CharlieBrown0115 Don't use Causvid. My current workflow uses Lightx2v which gives much better results and doesn't need 2 Ksamplers.
@hishiryo CAN YOU SHARE IT? im a bit dumb when istalling all the nodes a base models for each workflow!! would you mind tell me how to get the files from and where to install them ?>
I love what this wants to do, because none of the other LoRa's can manage this effect well.
I've found this very difficult to use as is. It seems like it could improve with a more racially diverse set of models and a broader set of expressions? As is, this tends to have a strong impact on the face structure and expression even at lower weights. If it blended better with wan cumshot, it would be an improvement, since it tends to be very responsive to specific prompts?
I hope you can manage more iteration on this.
Hi thanks for showing interest. I hope I can improve it more and I'm learning as I go. It's challenging to make it work with multiple sub-conceps and blending with other models. I have more ideas I want to try but it may get slow with the time I have.
I would really love to have this work the way I want too.
@boobkake22 By the way feel free to DM me resources if you know something that could fit my dataset in regards to the core concept. They are not easy to find with only two eyes.
I've noticed this as well. I really notice it messing with eyes.
@lethotrisv439 I'll message and see if I can be of any aid.
Damn this is really good. Hoping for a 720p version in future
Back to the model Wan Thiccum
Hello,I would like to express my frustration with the 'Wan Thiccum' model. I think it's very, very, very bad and doesn't associate well with other models. I think it would be better to avoid it.
I hope that this return will be taken into account for possible improvements.Thank you for your attention.
Chorsily,
I've trained a LORA that should work well with other CUMSHOT LORAs.
It can effectively solve issues like misdirected cumshots, shots not landing in your desired location, or semen appearing on the face even before the actual ejaculation (though these might still occur occasionally, for example, with very challenging angles). I've tested it, and while my LORA is primarily for anime styles, I've also found it works very well with realistic styles. If you'd like to give it a try, feel free to check it out. Wan CUMSHOT is truly a great LORA.
My LORA was trained using material created by Wan CUMSHOT.
My LORA:Anime Cumshot Aesthetics – Precision Load - Emotion Edition | Wan Video 14B i2v 480p LoRA | Civitai
If you prefer not to use my LORA, you can still refer to my previous article for guidance.
MY article:
How to use the suggestions regarding the Wan Cumshot series!! | Civitai
All my source material was created using the methods described in my article, but it required a lot of time spent adjusting weights and numerous generations to produce the video I wanted. That's why I eventually decided to train my own LORA to solve the time-consuming troubleshooting problem.
very cool but I want to see her swallow at the end, how can I prompt that?
Sorry but it doesn't know this concept
@lethotrisv439 I was able to do it somewhat, "she closes her mouth and tastes it" looks like a swallow
If the video stubbornly keeps making a new person come into frame or keeps shooting liquid from people's faces, mouths, or eyes, is that fixable?
I feel like it's the model because it's definitely not just this lora doing it, but maybe its an issue with many loras? Sometimes you can stack huge amounts of prompting to get it to stop, but by then the prompt is so big its probably ruining the video quality and even that doesn't work most of the time.
Is this an issue for others too? I think the most likely seeming issue is maybe it's because i don't make realistic videos and the loras are trained so hard on realistic videos that it just generates a new person since it can't figure out the drawing style. Some of these loras show non-realistic examples though, so that ends up just being more confusing :D
In a few of my generations I put "cum shooting from people's heads" or something along those lines in the negative prompt and it helped somewhat
I have also seen this a few times. When using image to video I want to think it could be something in the input image that triggers it. I don't know enough of the inner workings to be sure though.
so cool
do you have a workflow recommendation for that?
Best cum lora so far.
But can you make wan2.2 work? Only get blurred results
Thanks. Unfortunately it will be a while before I can spend more time on this, but I intend to.
Does it work with Wan2.2?
ok
we need this for wan2.2
I can't right now but soon
@lethotrisv439 hope top hear from you soon
while waiting for wan2.2 version, i tested this on wan2.2, it works, just put it on high and low strength 1.5
Works with Wan 2.2 I2V. I'm combining it with the f4c3spl4sh cumshot lora by K3NK for saucy results
How are you combining it exactly?
@engineer I just have the f4c3spl4sh loras (high / low) connected before this lora. I have them both at 1.0 but you can adjust each of them to your liking
@salsaverde do you connect it to both high/low? or only one of two
@salsaverde willing to share a workflow over?
@Hakazibo Yes, both high and low in two separate nodes
@bonypony don't know if you still need it but here you go https://gofile.io/d/op1dnS
@salsaverde thanks!
@salsaverde can you share the workflow again? the link died already lol
@wcaokc317 Sure thing https://gofile.io/d/KwQKUT
absolutely the best. doing anything cum related without this is like an apple pie with no apples. and IT WORKS WITH WAN 2.2! so everyone crying about it, download it and use it with wan 2.2, its great!
bro dropped one of the best wan lora and dipped
Details
Files
wan-thiccum-v3.safetensors
Mirrors
i2v_thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
147_wan-thiccum.safetensors
101_wan-thiccum.safetensors
wan-thiccum-v3.safetensors
i2v_thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
tc-v2-1860691.safetensors
tc-v2.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors
I2V_wan-thiccum-v3.safetensors
thiccum-v3.safetensors
wan-thiccum-v3.safetensors
wan-thiccum-v3.safetensors