Twerking LoRA for LTX 2.3
Trained on T2V but supports I2V
720x1280 resolution and 20-30 steps
Weight 0.8-1.0 is recommended
Prompt: she turns around, she starts twerking her booty and shaking her hips
⚠️ Try Sulphur if you have issues with the base model
Description
Initial first release for LTX 2.3
FAQ
Comments (20)
I always though squatting was mandatory for this
not in 2026
its an ai model bro. its not that deep bro. seek help bro
@famalam This is an intervention. You say bro too much. Everybody is here to help you and wants you to get better.
very good examples. however, I can't get the woman to move her butt side to side like you did in the woman with the horse example. how did you make her butt move side to side?
I think its the camera angle, you can also try adding 'side to side' into the prompt.
@AdaptiveVision i tried side to side, it still didn't work the way i wanted to... what camera angle do i need? also the video movement is also too slow...
@TypicalLoraLover. In my experience, the sampler you're using seems to make a big difference to how well it follows the lora and your prompt, and also whether you get an unwanted slow-motion effect. If you aren't already using it, try the "euler ancestral cfg pp" sampler. I seem to get a higher rate of vids at normal speed and actually doing the actions prompted for. It does run slower than some other samplers but is worth the trade-off, in my opinion
@benuk01 i cant seem to find the sampler in my program im using. im using wan2gp not comfyui for ltx 2.3...
@TypicalLoraLover There's your issue then.
@AdaptiveVision hmmm that's unfortunate. really wanted to use this lora... and i don't have enough vram or ram to use comfyui with ltx 2.3... oh well...
@TypicalLoraLover I'm using WanGP as well, they don't twerk at all, I just tried.
@AdaptiveVision I figured it out, i switched to comfyui! now it works great!
@TypicalLoraLover That's great to hear. I'm a huge WanGP fan for so many reasons but I've been getting better LTX2.3 results from some of the ComfyUI workflows - I strongly recommend the "Seed Hunter" workflows that have been posted recently where you can preview low-res generations (four at once) then proceed to full res with the ones that look promising. Also, the newer versions of ComfyUI seem to be much better at offloading to RAM if you don't have enough VRAM so people with lower spec PCs/laptops might find they can still use it successfully.
@benuk01 ohh this workflow here?: https://civitai.red/models/2676452/ltx-23-seed-hunter-multiroll-workflow
@TypicalLoraLover Are you using the Dev or Distilled version? GGUF or not? If you have a workflow, can you please share it?
@Vlonadio I'm using this one here https://civitai.red/models/2676452/ltx-23-seed-hunter-multiroll-workflow
Can you make a version for Minimax?
I was actually just starting to planning that, so yes 😊
@AdaptiveVision LESS GOOOOOO hahahaha I'm excited
Details
Files
trkti_lora_v1_600.safetensors
Mirrors
Twerking - LTX2.3 - she starts twerking her booty and shaking her hips.safetensors
Twerking - LTX2.3 - she starts twerking her booty and shaking her hips.safetensors
trkti_lora_v1_600.safetensors
trkti_lora_v1_600.safetensors
Twerking - LTX2.3 - she starts twerking her booty and shaking her hips.safetensors
Twerking - LTX2.3 - she starts twerking her booty and shaking her hips.safetensors