Strange Results
Everything seems to work fine, but the results I am getting are really weird. I haven't done any big changes to the workflow. However, the results I get are always the same.
Can anybody explain this to me?
ltxv-13b-i2v-base-fp8_gmbhneo.json
https://github.com/user-attachments/assets/7ef8484c-ddad-418c-8c00-d210677132c5
delete custom nodes ComfyUI-LTXVideo , clone again , use their example from ComfyUI-LTXVideo\example_workflows\13b-distilled
sounds strange but worked for me
I had the same results when I switched sampler from gradient estimation to some random one
I have this exact issue.. when I try the distilled version
For me, the "woolly knitting" effect seems to occur on various distilled models in i2v mode if the input image is too large forcing the tiling effect. Quite what "too large" means I don't know yet, but 512x512 seems to get things working well so you can at least get off the ground with 13b models.