Join the CoolStrad Community For More Stuff: https://discord.gg/RnUHPzTkXH
This is a LoRA that mainly adds a Digicam flash, blur, noise, and realism effect. (It can make sexy baddies as well 🤤)
This LoRA can't help with sexual acts, like missionary, doggy, blowjob, handjob, etc. You need to chain it with other sex based LoRA to make this effect work on those. Though it can make solo nude woman fine usually. Sometimes there could be vagina deformation, using any vagina LoRA along with this will fix it.
This is the Beta version for now, if I can further improve the LoRA, I'll upload that as well.
This is only a T2V and Ref2V LoRA, for image to video it wouldn't work as intended.
When using with Ref2V ensure that the weight is a little bit higher around 1.2, otherwise the effect isn't that noticeable. For T2V, 1.0 is fine, if you want the effect to be stronger you can try increasing it.
Description
FAQ
Comments (29)
If anyone has any questions, let me know here.
Thanks for making this! I think my main question is whether it affects faces overmuch, when used as references.
@feltmaterials993 Welcome! It can affect the face if the weight is high. But still, its pretty negligible. I tried making a video with my character on 1.5 weight. It made the clothing break, but the face was still intact with my reference image.
Fix your tag! You misspelled Minimax and it might be consequently harder to find.
@rkyha639 ah thanks so much, I didn't even notice I made a typo there. I'll fix it very soon.
Works amazing with reference at 1.2 str and the visuals are sooo good!
Hey, can you please share how you generate prompts for your videos??
I provide grok with the official prompt writing guide .md file, and I give it my idea and grok makes the prompt. That's exactly what I did for all of the showcase videos.
@coolstrad can you please share that .md file
@TaiLong a quick google search can help you, it's on minimax official prompt guide bro.
Sweet, I've been waiting for a lora like this. Thank you for sharing!
I'm a beginner, and your example video seems overexposed. Is the purpose of this LoRa feature to remove the oily skin (like a wax figure) from the original H3 model in the generated video? Thank you.
Not sure about the oily skin. But the point of this LoRA was to add noise, heavy flash effect if prompted, and increase the realism. Overexposed was intentionally prompted simulating the big flash of the camera.
Did you even read the description of this lora? WTF?
Bro can we talk if possible can you tell me how didi you trained the lora and on what datasets i am also trying to build one for myself , I am using aitoolkit
thanks for the lora.it can work successfully when use the fla models with the picture input.
this just seems to turn every woman asian?
Your social score -2000
If you don't add any woman based details then it generally makes the woman look asian. You simply need to prompt all the womans characteristics to not have the asian look. Another solution is to use the reference to video model with the lora. The lora doesnt destroy the reference characters face.
@gambikules858 why is that?
he is just being racist cause i'm asian
t2v = I2V same model. So why this model is for t2v and ref2v
It does work with I2V technically, the thing is that, this LoRA is just a visual effect, it wont do much with motion, as I2V continues the visual effect of the image, its not really worth it to use it with image to video, it wont work as intended. Thats the main idea. You can check what I mean if you run a test with image to video.
Your prompting is pretty detailed, something I do not expect anyone doing it here.. did you used LLM for that? if so could you share the instructions you send to it?
does it work with hybrid model?
I am not sure what the hybrid model is. I don't know if it'll work or not, didnt test it with that.
看到我們亞洲人太棒了!感謝你!!