UPDATE Nexus BTA My Web UI for Comfy with Predfined Workflow/template

https://redd.it/1tqq2zv
@rStableDiffusion

The best model for openpose / depth adherence

So I've been trying zit and flux2 Klein with control nets for depth and openpose and found the results pretty disappointing - they do alright with upright poses like a wave or walking from a side view, but when you try something more complex or upside down (flip, cartwheel) they pretty much suck.

Are all models suffering this same fate? They can only handle upright poses? Surely there are models even if they are old and clunky which has been trained better to handle all sorts of rotations of a person?

Any pointers to get better results? Workflows that seem to help? Your ideal setup of models?

https://redd.it/1tqxpzu
@rStableDiffusion

I Found it Real Easy to Make Your Own Character Lora Locally from Scratch.

https://redd.it/1tqz5y0
@rStableDiffusion

LTX 2.3 + LTX Director
https://youtu.be/_6w_S--2xB8

https://redd.it/1tr1xlg
@rStableDiffusion

Best local AI models for 16GB VRAM?
https://redd.it/1tr2ijg
@rStableDiffusion

Some Anima base generations

https://redd.it/1tr8xkx
@rStableDiffusion

Italy in the 1980s (Z-Image Turbo - Wan 2.2)

https://redd.it/1tr9539
@rStableDiffusion

Nava - A 6.3B audio-video model .

https://redd.it/1trb93v
@rStableDiffusion

Colored Noise Diffusion Sampling - plug-and-play, inference-time sampler.

https://redd.it/1tray25
@rStableDiffusion