Hi everyone! After a bit of fumbling and trial and error, I managed to scan food successfully, my goal with the Revopoint Pop. The depth scanner is good, and with a cheap lightbox I managed to get more out of the scanner than I thought I should.
With one exception, texture quality. It is bad. It’s too rushed. The point cloud extracted colors are often better than the photo textures themselves. I managed to make it a thousand times better with some complete retexturing, but it still leaves a lot to desire.
I think my gripes are also down to the fact that if the first time you pass on the object there’s a shadow, the shadows stay for good. Maybe with a suspended string technique or drilling into it + a stand I could get a cleaner texture, but the color pass needs improvement.
In a lot of Revopoint materials, you don’t see much color objects, or hear things like “not many users need textures”, or the fact the menus by default being on [No color] everytime… is a little admission that there isn’t much confidence on the texture side of the software.
Nonetheless, the scanner is good. With the right objects, it’s really practical.
I’ll post more in the future.
I used my phone to take pictures after the scan, with some diffuse light.
Shoved everything in Substance Painter, then photobashed it from every angle the 3d help me pinpoint. Then tweaked the texture into an albedo with Photoshop, and a few more maps there too (specular from albedo difference, cavity, micro normal from diffuse)
I’ve used that process before with 3d scan on professional AR projects, and recycling some knowledge here.
In Sketchfab It looks great but I see a substantial difference between the actual wireframe model and the model with surface mapping applied. Is the wireframe model all the geometry that the scanner captures or did you move some of the fine details of the geometry to the normal and roughness map?
If that simple wireframe is in fact the level of detail the scanner captures from the geometry of the object, How did you created the roughness and normal maps? from the photos?
@Artofwalls: Let’s try to explain my entire process. Ready for a long read?
Keep in mind, the idea was to have a useable game / cg asset, not a >1mil polys scan in the end. Scans are nice but useable topologies (and poly definition) is nicer for my job.
With that in mind, my process is always to have the highest details, polys and then bake it all on various degrees of cleaner/lighter meshes.
I’ll put here the final scan in zbrush, then under a quick zremesher. Would the scanner only have the low mesh resolution, I would be fuming (obviously).
I used the Substance Painter engine bakers to get to the lighter mesh the normals of the higher one. Makes it in Sketchfab so the textures load slower than the mesh but the mesh is there quicker.
#1 Diffuse = Complete paintover with the photo refs, turned into an albedo with Photoshop. There are bunch of tutorials of that online. Mostly duplicating the diffuse layer, desaturating it, then using it in soft light mode to ease the highlights of the texture, with extra tweaks. + Added AO + Added slight tweaks on the bone and blood.
#2 Roughness = Roughness was two things: Difference between Albedo and the original diffuse gives you a reflection map (that the albedo conversion removes) + I also cranked up the highlights of the ice and frost white bits to add them back to the roughness.
#3 Normals (see below) = High poly scan normals baked in Substance + later along the process, a trick I learn working for BBC 3d scans: in Photoshop, turning the diffuse in a Normal Map and adding that too on top of the object for extra extra details. The scanner wouldn’t have got the very thin skin folds on the leg. + Some height from my frost/ice mask.
#4/5 Handpainted SSS/Scatter map using my gut feeling +a thickness map -the bone and some cavity map.
NB: I might be able to achieve all that using Substance Alchemist, but haven’t tried this route yet.
