You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Trellis on Hugging face generates the video to the side, downloading this will give you a usable video plane(card) of your object from many angles, this made me think of adobes 3d rotation tool, which will redraw your art to a new angle. Current methods of using Ai tend towards displacement for smaller detail. Assuming trellis is context sensitive and could generate depth based on points of objects, which it seems to be. Could we not use planes with vdm textures from the already made albedo, normal map, and a newly made vector depth map?
This would simply give us the ability to generate details at different angles on the fly, Say An air conditioner seen from the front, oh give it a lil turn, u still get full texture quality instead of blurred lines down.
Here Are a couple pictures with my trials using Simple Displacement, from a video file, much like openexr for VDM.
With this yes you'd not have the whole model, but it wouldn't matter as if implemented like adobes you could simple turn to have the angle you need, and spend half the amount of geometry as the inverse is automatically occluded. Plus Higher detail video generation would work better as youd just scroll in to get more, much like photogrammetry or lods.. Just a thought Let me know what yall think! I think if implemented like Adobes in blender, then It would very much help my process. So Id like to know its viability.
The text was updated successfully, but these errors were encountered:
Trellis on Hugging face generates the video to the side, downloading this will give you a usable video plane(card) of your object from many angles, this made me think of adobes 3d rotation tool, which will redraw your art to a new angle. Current methods of using Ai tend towards displacement for smaller detail. Assuming trellis is context sensitive and could generate depth based on points of objects, which it seems to be. Could we not use planes with vdm textures from the already made albedo, normal map, and a newly made vector depth map?
This would simply give us the ability to generate details at different angles on the fly, Say An air conditioner seen from the front, oh give it a lil turn, u still get full texture quality instead of blurred lines down.
Here Are a couple pictures with my trials using Simple Displacement, from a video file, much like openexr for VDM.
With this yes you'd not have the whole model, but it wouldn't matter as if implemented like adobes you could simple turn to have the angle you need, and spend half the amount of geometry as the inverse is automatically occluded. Plus Higher detail video generation would work better as youd just scroll in to get more, much like photogrammetry or lods.. Just a thought Let me know what yall think! I think if implemented like Adobes in blender, then It would very much help my process. So Id like to know its viability.
The text was updated successfully, but these errors were encountered: