Strength range : 0.5/1.5
No specific trigger words
The dataset is huge and composed of images and videos
Like with Flux or any other diffusion model, stacking multiple LoRAs often leads to artifacts and degraded image quality.
All videos preview are rendered at 8 steps with causvid LoRA v2 without any other LoRA applied
Description
FAQ
Comments (31)
Why is it labeled as Trained Checkpoint?
Nice catch, I changed it thanks
Great Lora!
Thanks a lot!
Any prompt examples you can share?
I use basic Florence2 generated prompts, and change "still image" related words for "video" related words, adding (dynamic, movements, motion, realtime) things like that
The "flux face" is strong with this one.
No flux faces has been trained
pgc 100%, there are absolutely images generated with Flux in your dataset. The butt chin appears at times and the Flux aesthetic quality comes out sometimes as well. Not all the time and I am not saying it is a bad thing. I am just pointing it out.
playtime_ai There is about 15% of Flux images for body shape, with white background and low cfg to have raw looking images, but no faces from it, I usually crop them to avoid having any faces that can feed the model with the same face pattern that flux have, this is a part of the faces and eyes dataset, there is no generated images I can guarantee
Is there i2v or just t2v?
It's trained on T2V but it should works with I2V
I plugged this in to a WAN workflow without a clue about optimizing settings. Of course, the results were abominable. A hint would be appreciated. All the gallery examples are posted without prompts or settings so no help there.
I posted a simple T2V workflow on the download section of the model, I hope this will help you
@dcham2310 you can download the videos and drag them into comfy to see the workflow.
@Griphen116 I may use some nodes that I didn't push on github yet, but yes you can do that
Thanks, I didn't remember the thing about dragging the video. The workflow will be welcome though. I work more with i2v than T2V but I'm guessing that will be ok too.
seems to enhance NSFW movement very well , really like what it does , but I'm finding at higher strengths, like over 0.6 it blows out the image making it noisy , I am using it for I2V with lightx lora so maybe that's why , know anything i might try to get the vid sharp again
I was skeptical of this lora given its lack of usage instructions but it singlehandedly saved my use of a character lora + M4CROM4STI4 + pussyjob "v2 - alpha". Without this lora, this combination couldn't retain the correct look while keeping the pussyjob motion.
Because this lora naturally raises breast size (at least when combined with my huge breast size prompts), I was able to turn down M4CROM4STI4 and recover the pussyjob motion while retaining the correct look. For my application, 0.4 was the right strength. Any higher and it changed the pussyjob to cowgirl.
it's good, but i wish it wasn't all giant nipps
I'm not a NSFW master at all. I like this LoRa but it creates ginormous breasts. Visually everything works but the "average" breast form this LoRa are substantially larger than that blonde that is on that show and all the movies that has the "best boobs in Hollywood."
Maybe try the "boob size slider" or "Small breast Helper" LoRA(s) in addition to this?
It's amazing man
Missing this Lora in Wan 2.2! Any chance for an update?
I'm training it since two weeks now, this will be a huge one
@pgc Glorious. Thank you!
Just wondering if it can be used with wan 2.2 as well, even if it is not trained with it. I'm learning this generative AI, and I'm using LORAS of wan 2.2, 2.1, I2V and T2V indistinctly. Sometimes weird things happen, but usually it's all right.
@pgc Any ETA on 2.2 version?, thanks!
this is the only lora i use. this is the only lora i need. please add a wan 2.2 version of this lora
I use this for 2.2 and works better than any of the others, would love a 2.2 version.
Hi can you please explain how to get the workflow using a video dowloaded, it doesnt work on my comfyui, TX
This is a LoRA that is used within a workflow (typically called a JSON). Within the workflow, there are a number of nodes, and one of those nodes is to enter various LoRA. This LoRA helps generate the videos based on your prompt and other criteria.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.