Hi all! lastly I’ve been learning and diving with ComfyUI, looking to add new solutions to my pipeline, as new times come ahead. I’ve already posted a first part, about normal and depth extraction with comfyUI, and using that as part of a Vfx pipeline in Flame (sadly facemaps went oldish), and my idea is to keep posting and sharing all that I can get and learn. I’m trying to learn now how to get more passes (albedo, roughness, specular etc), any hints welcome
Next week I’ll post another video combining in Flame a camera track from Syntheyes, and a generated model, coming from hy imgto3d (working on it right now). If anyone wants to add some ideas, tools or just to ask about any detail, I’m here!
You can check the 1st one here, I dont dive too much into details as I wanted it short and the less boring possible but the concepts are there
Great thanks for starting this, I was gonna do the same.
I’ll add some proper stuff at some point. But let me start with this little tip. Comfy UI is setup for use with a mouse, but if, like me, you’re using a tablet, then it can be a bit tricky to navigate without constantly using the mouse control wheel. However in the main schematic view, if you press shift = ctrl and drag up and down you can zoom in and out of the schematic like in Flame. Useful, quick way to navigate for us tablet heads.
if i remember correct resolution wise the new beeble studio ( local installation ) is 4k with a decent gpu (over 16g)
the results are ok-ish might need to retrain it with Thiagos tool . Talking about the inconsistencies temporaly . if you have a test footage send it over and will export it out and you can have a look
this is a nice one showing some things about normals ,Uni3C, ATI and he shares his json
and this for upscaling locally . it has parameters to make the model work with mid range gpus.
