stabilityai

/

stable-diffusion-2-depth

like 370

This guy on Twitter is using depth map from Blender, maybe you can contact him for help -> https://twitter.com/Nymarius_/status/1598086899922391040

Edit: + https://www.reddit.com/r/StableDiffusion/comments/zh8ava/using_depth2image_with_detailed_depth_maps_from/

+1 for this

• Depth can take significant time for some images, so calculating depth once and generate SD variations faster
• 3D software: it seems better to have depth map as starting point (really fast to render) than trying to get a simple untextured render to convert into depth and then making SD variations

+1 makes no sense if I cannot, no sense at all
also where is the generated depth?

I found the origin
https://huggingface.co/spaces/nielsr/dpt-depth-estimation

I'm creating a colab to use custom
Open to collaborators!

https://github.com/albertotrunk/depth2video

there is an undocumented depth_map (expects a tensor, convert it from a PIL image) argument for pipe of the depth2img pipeline in diffusers