So I was revisiting my old stills, looking for material I could animate, and I realized that my old stills were sexier than my recent videos. At least in my mind. Even the hunyuan video work was sexier in some instances than the WAN video stuff.
I don't know if they have an option to do a poll, but I would like one. I am curious, do you guys like the Hunyuan Video stuff (more natural behaviour, big boobs possible, some anomalous warping, very short videos I posted around page 26 for a few pages), the Stable Diffusion stuff (every still early on in this thread up to around page 42 or so was Stable Diffusion plus the crappy animated gifs up to page 26), the Perchance stuff (pages 36 to 45), the Pony stuff (stills page 45 to the end), or the WAN stuff (videos page 47 to the end).
For a side announcement, I combed through MacromastiaDream's images and videos and was able to pull out enough metadata to determine he is running a kijai version of WAN with some of the same LoRAs I use. I don't get the results they do, but I am potentially determined to try. Also, when I asked Grok what the most popular AI video generation software was based on available checkpoints, downloads, and LoRAs, it came back, almost definitively, that WAN is the clear winner. None of the other models come close on community support. CivitAI.com has tons of WAN LoRAs for almost anything you could imagine.
Anyway, I keep trying. I have streamlined my build to where I have one ComfyUI install for WAN and one Auto1111 install for Pony and that is it. They are stable, they work reliably, they both have tons of LoRAs available, and they give adequately acceptable results. About 250GB of my drive is dedicated to AI models and LoRAs, which is a lot less than it used to be. If I ever decide to stop doing this I will post a tutorial on how to install ComfyUI with phr00t's rapid all-in-one mega checkpoint, the workflow I use, how to use LoRAs, and then also do the same for installing Auto1111 with the CyberPony checkpoint and LoRAs.
Punchy