RuneXX

/

LTX-2-Workflows

like 271

Try start the prompt with something along : "Make this image come alive with fluid motion..."
And focus the prompting on sequence of actions, and less describing looks ...

"first he does this, then this happens... and then ... and then..."

But yes, some images can be stubborn. Thats why there is an image compression node in there as well. To help the model think of the image as a video frame, and not a still..

Try start the prompt with something along : "Make this image come alive with fluid motion..."
And focus the prompting on sequence of actions, and less describing looks ...

"first he does this, then this happens... and then ... and then..."

But yes, some images can be stubborn. Thats why there is an image compression node in there as well. To help the model think of the image as a video frame, and not a still..

I used all above, even compression 42 lol. I'm lucky prolly... I hoped there is a simple solution 😂
TY

If you get a super stubborn image, you can also use the image2video lora, that seems to help.
(but i havent tried this with LTX-2.3, it was made for LTX-2.0... i'll give it a try when i run into a stubborn image)

But if you want to try it :
https://huggingface.co/MachineDelusions/LTX-2_Image2Video_Adapter_LoRa/tree/main

If you get a super stubborn image, you can also use the image2video lora, that seems to help.
(but i havent tried this with LTX-2.3, it was made for LTX-2.0... i'll give it a try when i run into a stubborn image)

But if you want to try it :
https://huggingface.co/MachineDelusions/LTX-2_Image2Video_Adapter_LoRa/tree/main

Yes. It works with 2.3 too. So no easy way 😂
Thank you for your WFs, btw

Most welcome ;-)

I havent run into stubborn images so often, hopefully you wont either ;-)

Most welcome ;-)

I havent run into stubborn images so often, hopefully you wont either ;-)

Lol. Not a single still image. Just the talk was enough

Wardrobe malfunction ;-)
And possessed by the devil haha