So you mean that RC is doing a bad job at simplification (RC introduces a ramp (where I drew the red circle)), and the blurry textures are because of that?
If yes, is there no way to fix that in RC? I would like to stay in RC for the entire process if possible.
Here is a screenshot that shows the wall from the side. I don’t think the “ramp” is so much “angled” that it would explain the blurriness. What do you think? In fact, the geometry looks rather nice.
I would bet my left hand that this is the cause! 
If you look at one of your comparisons, it’s exactly the areas that deviate greatly from the original where it gets blurry.
You can’t blame RC with such an extreme simplification. There are some tools that would do a slightly better job, but there is a limit. Why don’t you try 10k and see where that gets you?
I don’t think RC did a bad job on the simplification.
It just confuses me that the textures in this area are blurry.
This doesn’t look understandable for me.
Could this be a bug?
Or perhaps a setting that just isn’t right?
The geometry looks really nice, I don’t understand the blurriness.
No, not a bug. Definitely!
Why don’t you export both models and load them both in CloudCompare or MeshLab and look at the differences.
And DO try the 10k model - it will look better!
10k does look a bit better but still blurred.
What I don’t understand is:
The geometry of the real object and the model (full res) and the model (10k or 1k simplified) is nearly flat in the area that I drew a red circle on.
It isn’t so that the real object is highly detailled while the simplified version is much less detailled.
However, after simplifying and unwrapped and re-texturing, the blurriness occurs.
Why is that? What has changed?
If you now do a 50 or 100k and those look even better, then there is a definite tendency…
This is all I can come up with.
Would be nice if you could upload the 10k (and following) in textured and solid view…
This is the 10k model.
I don’t see much geometry change in the region that was blurry before.
So my question would be: Why does a little change in geometry result in such a big difference (blurry texture vs. sharp texture)?
And the next question would be: How can I work around this problem?
Can the developers of RC also tell their opinion here?
For completeness reasons, here is the 33mio tris model in solid mode:
As one can see, both the 1k model and the 33mio tris model have a nearly flat surface. That’s why I don’t understand why the 1k model is blurry.
Shouldn’t RC be able to reproject the texture correctly?
Here is the image set: https://drive.google.com/open?id=12UVqTFWFrCVT0uKwjW1dsUNk8bwIiQOy
try taking the full res textured version. and the low res model to xnormal. and reproject textures there.
you probably also want to manually unwrap at least the low res version first.
keep in mind you need to export the high-res version as one texture.
Unfortunately I’m not experienced enough to do that.
Dear Alice,
your issue is probably caused by the large simplification of the model, as has already been said. Large simplification of the surface causes small, yet significant movements of the surface which leads to discrepancies between the textures. Based on our experience, we generally recommend to simplify models 10 times at most to preserve the detail which does not cause problems in the texture.
Hi alice,
wit you saying that the original model was 33m, all is said!
Even your 10k model is at 1/3300 of the original size, way beyond anything that I’ve ever seen. Look at the edges of the wall or the cigarette stump for example.
As lubenko explained, 1/10 is a rough guideline for a mostly loss-free simplification.
I do much more than that (from a few hundred m to usually around 10m, but it highly depends on the object and the purpose for which the model is intended. For example if you take pictures at always the same distance of a very uniform but complex object (e.g. a plant), you will notice a strong simplification sooner than a relatively simple object (building exterior) where you also took images at very different distances. The reason is that in the latter, a lot of detail from the close-ups can fall to the wayside without influencing your intended level of detail.
More triangle count is not possible for me as I create game models.
Even 10 k is too much for this object.
So there is no other solution than using an external program, and the current solution for this is the final solution and it will not be improved?
Can you not integrate InstantMeshes?
I know Modo and Remake use it.
Or how could I use it?
When I have run InstantMeshes, and I have a nice UV unwrap, how could I bring that to RC?
Hi alice,
if you have created a low poly mesh with external software, you can import it.
There are quite a few threads dealing with that, so I am sure you’ll find what you need.
Just bear in mind that RC can only process tri meshes.
I would love for the guys from Valve to create a course just on this issue. I would gladly pay for it!
I am running into a similar problem but slightly different, I create a low poly flat surface for simple surfaces in Modo but running into a different problem with it texturing looking odd and not lining up with the high poly mesh perfectly.
P.S. Gabe and friends if you are listening please put together a class on this 