Change the way you approach LTX genning. We all know that prompt adherence with LTX is notoriously bad. The model treats your prompts more like suggestions that it can ignore at will. So I say, stop beating your head on the wall trying to prompt better. Instead, the path to success with the model is to see the outcomes of many seeds quickly, and be able to choose one to bring to finalization. This is the main workflow I use to produce my short films, and I hope you all enjoy it too.
See my full tutorial video showing how to use the workflow here:
(Note that this video shows the older 3-Stage version, but usage is more or less the same)
Description
Adds Eros v1.4 support, RIFE Frame Interpolation, and cleans up the overall flow!
FAQ
Comments (2)
Thanks for this, it's working well so far. It is called LTX/EROS, so which eros model have you been using? Or have I misunderstood what EROS means? So far, if I try anything other than LTX Distilled 1.1 transformer, i get grainy mush. The workflow you provided used an ltx distilled convrot.
Yeah, int8 is great for 30XX/40XX cards, and I have a 3090. If you look at the note above the model loading subgraph, I link to the version of Eros I've tested with the workflow (1.4 DMD turbo). All the links for the various models are there already for you to copy/paste.