Just updating my adventures with pc based AI tools. Have a really good understanding of sdwebui now, been using it to make deforum animations, testing qr code images, controlnet, etc (on sleeplessmonk.gumroad.com you can find a massive one hour vj pack I recently made with deforum). It works good on my 4Gb VRAM with the latest update, I could make animations in SDXL and SD 1.5 at 720p. Still some bugs and some plugins donot work with each other. I really wanted to work with animatediff prompt travel, possibly the most advanced AI video method that can produce very realistic VJ loops and cinemtic content (it is similar but far more advanced to runway, pika video models), but I simply could not get it working in sdwebui
https://github.com/continue-revolution/sd-webui-animatediff
This got me back to comfyui. As an update now comfyui has some very interesting video workflows implemented from deforum. Everything is super easy to install and it has a portable version (which resolves extremely annoying bugs when you work with python on an external drive, this was super annoying with the sdwebui). As it is node based and errors are rarely encountered it is more user friendly, true the node workflow can be intimidating for those who have never used touchdesigner, vvvv or the likes. Below are the steps for implementing animatediff in comfyui, super easy and non-coder friendly. Overall there are pros and cons to both sdwebui and comfyui and I'm getting used to both. I'll post some first animations in the comments over the days.
Discussion (1)
This is a game changer. I have been slowly trawling through tutorials. The future is bright, I hope!