Am I doing something wrong here....?
So I've tried the new T2V High and Low loras, and the results are consistently a lot worse than even the previous versions... I've tried Kijai's versions of them and the same holds true... I've tried using the Dyno high noise model with the new low noise lora and I'm still not getting good results...
Is anyone else getting these to work well? I'm trying to work out at this point if it's something on my end or if it's a general issue...
I've tried using the workflow embedded into the video examples as well as modifying the previous Comfy Native workflow by changing the models and the loras... If anyone has got them working well, I'd value your guidance here!
I've discovered that the issue is largely related to the prompt wording. Videos generated using the example prompt exhibit normal motion, but when I use the prompt I created for training my own LORA, it doesn't work.
Interesting - so the LoRA needs very specific directions to get it to work otherwise it falls apart when left to its own devices? Hmm...
Interesting - so the LoRA needs very specific directions to get it to work otherwise it falls apart when left to its own devices? Hmm...
Hahaha, my English isn't great, I used a translation app. What I meant was that its movement is very slow.