Question about workflow combination of nodes. Win11Pro with MSI Intel Pro processor. 64GB Ram, RTX5090 32GB Vram so I should be good running FP8 model. I’ve been trying to experiment with LTX2.3 video I2V. CUI’s workflow (as well as some others I’m trying) for the LTX2.3 I2V template uses a subgraph so all the ckpt, loras, text enc, upscaler, etc. are packaged on the front node. I made sure all the models were exactly the ones expected; that’s fine. However, in running just trying to run a test with my supplied photo and the text already there, it runs (without complaining about model choices) until such time the computer crashes (OOM not doubt!) only a 640 x 480 testing attempt. Done so repeatedly, will not complete.
I’ve de-subgraphed the wkflw so I could change to a GGUF model to see if that worked, but that node only has “model” and no clip or vae for connection. UGH!
It seems I should be fine with my system specs, though, even without GGUF. Any suggestions/experiences to point me in the right direction? Thanks.