Well, thanks anyway!
Ever since this model dropped, I hoped you would find a way to trim the fat off that humongous atrocity of that workflow and make a single file. I think it's gotta be possible, even if the "simple workflow" ends up hiding a bunch of collapsed node groups.
In the meantime it seems WanGP supports it? But I'll still look forward to whatever inroads you manage as for merging popular LoRAs and whatnot into a does-it-all model.
It is just a mess to try and package as an "all in one". The text encoder is really big, and it requires another "embedding" model to be loaded with a DualClipLoader. Then you've got 2 VAE's and the "Load Checkpoint" can only load 1 VAE. Ugh.
The official LTX2 workflows do hide a ton of stuff in a subgraph, but you can't just use a modified diffusion model (like mine) with it, because it expects the video VAE to be included. You could use it, but it requires unpacking the subgraph and making changes, which isn't really simple anymore.
If I figure something out, I'll see what I can do. I'm settling now for just tweaking the diffusion model, which is much more straightforward to distribute, but doesn't help the spaghetti mess LTX2 usage is.