LTX-2.3 Update Release!
• Fully compatible with LTX-2.3 I2V pipeline
• Audio generation included (real synced sound!)
• Much improved growing/morphing quality & detail stability
• Cleaner fur patterns, better anatomy during TF
FurryTF LoRA I2V Weight from LTX-2.3: 0.8 - 1.0
[Options] Furry_Realistic LoRA: 0.8 - 1.0
New Update on V1.1b
• step was upgraded to 4000 steps
• quality has been improved (Maybe.)
So improvement quality comes later...
Important Instructions:
You need to use both versions: high noise and low noise.
If you try to use only the low noise version of one model, it will not work.
For the Plus (Options):
You can use the “Furry Enhancer Video LoRA”, but only with the low noise setting to improve realism.
If you use it with high noise, the result will be a black video, which is not usable.
✅ The safest approach is to use the low noise version of Furry Enhancer, and everything should work fine.
FurryTF LoRA I2V Weight:
High noise: 0.8-1.0
Low Noise: 0.8-1.0
(Options) Furry Enhancer Video LoRA I2V Weight:
High noise: ❌ (This results in a black video image.)
Low noise: ✅ 0.8-1.0 (for Realistic or maybe cartoon else.)
Description
FAQ
Comments (9)
can you do gender body transformation lora.
you can do that too.
Wait! No
I have a request/recommendation for this concept.
I assume this is intended for first-frame i2v, and then they turn into the animal you specify?
Could you train this in reverse, starting with an animal that becomes human, and then we can reverse the resulting video ourselves? Because for a lot of people, the final result is the most important, and the human you start with not as much. It would let you get all the details of the creature to transform into correct when it's your starting frame, and then turns into a generic man or woman (which, again, the point is to reverse the video after generation).
Okay and what do you mean by that?
@Mistermango23 I'm not sure what you mean. I'm saying that if we can generate a first-frame-based i2v video of a dog turning into a man, we could then reverse it manually in a video editing program. People who want to see transformation are more interested in the end result than the start, so giving control over the "start" like this would possibly be more helpful.
Play the generated video in reverse... (Avidemux, Premiere Pro, KDenlive, etc.)
@Mistermango23 I think he means train a anthro to human lora. Giving the user more control over the furry character which usually matters more in this scenario. Or if its alright you can share your dataset and i can try to create that lora.
@Anthrats @SD_AI_2025 @CyberChili Guys, that's the reason why I'm creating a lora for creation? Because I enjoy training loras, and people enjoy it when I do it.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.