Since your model is for doing Image 2 Video then you should select the I2V type resources of this lora. The different available resources are listed in a row at the top of this page. You'd needs the two I2Vs (High and Low)

@c0ur4ge I just said if I can use t2v for i2v since t2v has a most recent update than i2v one.

@Alright123 Yes it works quite decently for I2V as well

@c0ur4ge For the I2V set, how the heck do you maintain faces? The lora is trying to change anime faces into IRL.

yeah this lora is too easy to get lost in a collection of better named loras

@c0ur4ge You should be able to delete the file from the version and replace it, it won't nuke the versions automatically or anything, it will just show a warning.

@c0ur4ge yep I know how that feels authoring my own Wan motion loras - it's an absolute beast of a task! But we get there through determination, intelligence and pure grit! lol

The lora is probably very good, but why do the creators never publish the minimum workflow for launching the model when they release it? Why is there no information about the strenght level to set for high and low lora's? My questions about I2v

@maxymirnow not sure why you need the workflow info. It's just a LoRA, and you incorporate it into your ComfyUI/SwarmUI/Automatic1111 workflow using load lora nodes.

Ideal strength info can be helpful though.

@c0ur4ge Where? Where is the recommended resource section?

Sometimes I feel like these people are looking at this from another website, recommended resources? where the hell is that supposed to be at?

@keedspiller49552 On desktop scoll all the way down, below the Gallery is the "Suggested Resources" section. On mobile and on other pages it is sometimes above the Discussion section, no idea why they have different layouts...

@javbukkake Sharing is caring bro. It is not that diffucult

@zirtapod Totally agree, never said the opposite: sharing is caring. I was just sharing my personal experience: I tried ComfyUI countless times using others workflows and abandoned each time, going back to Pinokio and Forge. Then someday, with the help of Gemini 2.5, I finally! decided to understand each node I was using, one by one. And it's only doing that, and creating my own workflow step by step, understanding the why and how of each node, that I finally abandoned all other UIs and moved all my AI workflows to ComfyUI. What I tried to say was: using your brain rather than count on others' intellect, is the only way to fully understand what a "workflow" means.

@javbukkake You're absolutely right. Using someone else's workflow either lacks nodes, lacks sageattn, or there's insufficient graphics card capacity. I really don't know where to make modifications.

Check out taggui for captioning:
https://github.com/jhc13/taggui

@ERA5ER Thank you. Is this a video tagging tool or an image tagging tool, or does it support both?

@wange999 That just does images. I've been using TripleX to extract single frames from videos, caption those frames with taggui, then refine the captions a bit by hand to include descriptions of the motion. Then pair those txt files with the video files for training.
https://github.com/NSFW-API/TripleX

It's totally capable of doing that, I have no issue. You just have to describe it in your prompt. Instead of saying "deep throat" say something descriptive like "she moves her mouth up and down his penis in a rhythmic fashion"

On her pics 😎

"I build small neural networks for the depiction of things that the larger neural network cannot" sounds way better than, "I make funny foot porn."

On like resumes and CVs.

Do not confuse "lora not work" with "I tried to make it work, I failed, it has to be someone else's fault".

how'd you go about using both smash cut and dreamly to do that?

@xavierlavion111 I only used smash cut to steal it's final frame, but then it was all dr34ml4y, just using a smash cut like prompt. It was the previous version though, this one they both try to move to the new position. https://civitai.com/posts/24126573

@c0ur4ge what sampler should be used? Ive tried unipc and eular, and dpm++ and still have visual blurry squiggly weirdness.
I use the lightning lora, but no other lora had issues with it.

@S65 same,
you figured out the problem?

@pankerf32448 nope

I encountered the same issue. I've been experimenting with adjusting the weights of the CFG and LORA parameters. I discovered that when both CFG and LORA weights are set to 1, the filter disappears. Properly balancing their parameters might resolve this problem.

try setting it to euler then select "simple" on the option below it

@charlenebelmont im using wan2gp, don't see a simple option, just the sampler option for eular

@S65 then this might be a case of the ksampler . test on a different work flow first. the best option is alway start of with the basic. testing it and seeing how the model does before changing anything. any custom workflow is set up a way that might not go well with the model you are testing.

Please! I¡ll buy you a lot of coffees :)

@c0ur4ge by far the best nsfw lora

you can in a lot of cases actually format prompt like this:

(at 0 seconds: 'yourprompthere')
(at 2 seconds: the camera shot changes to 'yourprompthere')

the wan22EnhancedNSFWCameraPrompt models that came out recently are especially good at this kind of stuff, but for total cuts like this, face and body preservation may vary. It can be better to (if you dont want to make a new source image) make a video of them switing positions that is seamless, and then just using a late frame in that to make the next shot, delete the video that had the transition, and splice the two shots together

strange, works perfectly fine for me (wan2gp)

works just fine with anime

I'm still learning also and what I have been doing is using the word "immediately" in the prompt with good success: the woman immediately turns, the woman immediately sucks, etc....

@bigconcat these loras are confusing. it's tough to get the strengths right too

@takearushfan374 My solution to that is to save an image/video uploaded from one of the users and drag into ComfyUI. Then look to see what settings they used and maybe just use their workflow too :)

@bigconcat emmm.....For example, how can I find the workflow on this LoRa page? I tried saving some sample images as JSON files, but that didn't work. How can I tell which images contain the workflow?

@zhle1201540 It's a hit or miss if the workflow will be embedded. You save the image with its original extension, then drag into Comfy. If the image is a png, its more likely to have the workflow embedded, same for video. jpg can work also, If not, at least we can see the prompts/lora/model's they used on the right side of the page when you click on the image here.

@bigconcat For a newbie this all gets confusing. Combining loras/strengths/prompts. Heck I'm still trying to find a d-ck that isn't uncut lol

dark ebony skin

@c0ur4ge Would it be possible to get some training into the next version for more saggy/Tuberous breasts?

@c0ur4ge Shame, cause you could have made bank back then before the payment processor BS that made every site worthless. Thanks anyways. Because the tags aren't actual words, they trigger the LoRAs perfectly on there.

Imagine getting one of the best lora out there for FREE and somehow still finding something to complain about. Wild.

@c0ur4ge Dont stress over it brother. Keep up the great work, i love your stuff!

@c0ur4ge Look how many upvotes my comment is getting. CLEARLY im not the only one with this issue!

@c0ur4ge i simply can't trigger that neither,been trying for 2 weeks with over 40 differents lora(not all just for that tho)haha, different prompt,diffenrent cfg, different resolution and NADA! :( . but yeah possible with t2v t2i but i2v is imply impossible unless you have a starting image with action already there. if there is any suggestion i'll take it lol! and btw i second Antisane, your lora is simply the best out there after all theses testing i've done! thanks a lot!

@c0ur4ge If you can get this working as it is now (you imply that you can), could you give us an example prompt?

@c0ur4ge really appreciate your work! if you find a way to fix that i promise to make a donation! haha!

Hi, can't see anything in your profile. Can you post some examples here please?

@prabhjot5877 how about this? https://civitai.com/posts/25307882

@BlatentLatent nothing either

@logicalpop that's weird, this is the link to the image post https://civitai.com/images/114684564 it has reactions so I know other people can see it, not sure if it's a civitai issue or what

@BlatentLatent got it now! must have been the site, had problem with it lately! thanks! :)

Work perfectly for you?

@p3ter_ai Yes! I use it in basically every video.

@StrangeLimina I'm using a comfyui workflow template, but not always it doesn't always adhere well to the prompt. What workflow do you use?

@p3ter_ai I use a custom workflow that uses stuff from a couple different workflows patched together. Mostly it's the Smooth Workflow from the SmoothMix guy:
https://civitai.com/models/1847730?modelVersionId=2110988

For prompts I suggest scrolling down to look at examples posted by others, finding one that looks like what you want your video to look like, and copying/modifying/mixing prompts experimentally to see what works together.

Are you using lightx2v?

@WetFeverDream yes, should I turn it off?

@limbojunkie No, I just needed to know to give you better advice. So if you increase the weight of your high noise lightx2v lora, the broad movement will be faster and the scene is inclined to change more. If you increase the weight of your low noise lightx2v lora you will se more more of the details you specified in your prompt as well as faster movements in those details. The high noise lora will make your car go faster down the highway, the low noise lora will make the wheels spin. High noise lora with low weights + low noise lora with high weights will give you a stopped car with shaking wheels. Basically, if you need more movement try increasing the weight of the lightx2v loras. Prompt is also really important. Both the right words and the length of the prompt will impact the speed.

@WetFeverDream thanks a lot!

@WetFeverDream Thank you for sharing this!!