So I want to create a 3D asset. I want to bake my highpoly mesh to lowpoly. After baking, I tried the baked result to lowpoly, but why does the result have edges like this in the lowpoly mesh?
Not enough space (in blender, as this seems to be, called margin) between UV-islands ?? Just a guess.. you might give some more info.. Also the low-poly seems to have orthogonal but not-connected faces.. and n-gons.. (even if that part seems to lokk okay).
Not enough space (in blender, as this seems to be, called margin) between UV-islands ?? Just a guess.. you might give some more info.. Also the low-poly seems to have orthogonal but not-connected faces.. and n-gons.. (even if that part seems to lokk okay).
I set my UV padding to 16px and I tried to increase the padding to 32px to bake in 4k resolution textures
Basically yes , because its already a seam for a videocard and every hard edge is a split edge doubling vertexes along it. Yet not every UV seam needs to be a hard edge and you don't have to always do hard edges along beveled edges. Google and try to understand face weighted vertex/split normals or ask chat GPT?
Shading of any polygonal surfaces is defined by vertex or split (in Blender terminology) normals ( pink vectors) first. Normal map is just a cherry on top of it. Once you figure out the concept of vertex normals and how they create shading the else become instantly obvious.
@Noanoww "okay, so every hard edge has to have seams?"
To understand it fully, simply observe the current normalmap of this area (and also post a screenshot of it here, for the sake of future people potentially viewing this thread) :
The cause of these artefacts will be pretty much self-evident once you look at it closely.
@Noanoww "okay, so every hard edge has to have seams?"
To understand it fully, simply observe the current normalmap of this area (and also post a screenshot of it here, for the sake of future people potentially viewing this thread) :
The cause of these artefacts will be pretty much self-evident once you look at it closely.
Hi, I am suggesting you to post a *screenshot of the normalmap* corresponding to this area. Once you do so and take the time to observe it closely, you'll understand your issue.
Also, @gnoop , I hate to be a bit of an ass but "Google and try to understand face weighted vertex/split normals or ask chat GPT?" or "something something videocard" isn't contributing anything to help the OP understand the issue. If anything this is just adding even more misinformation/slop to a topic that has been hard to grasp for artists for years.
Select the object then look for the "UV Editing" workspace tab at the top. Press "A" to select all geometry. We want to see the UVs that'll appear on the left.
In case there's no image under the UV you can select it from the header in the UV Editor or click on it from a Shader Editor and it'll appear automatically, but in this case having the normal map itself visible isn't necessary to see how the object was unwrapped; you'll see it highlighted as geometry on the left if you selected the vertices (or use UV Sync Selection).
isn't contributing anything to help the OP understand the issue.
Yeah, Google maybe is not very specific and helpful advice but chat GPT sure is. I didn't ask it to explain me vertex based shading personally but did for a lot of real time shader and pixel manipulating techniques and it have been best teacher and best focused and compacted info no nvidua published GPU gems ever been for me .
the robot is only going to recycle the misinformation this is a subject that cannot be googled because at least half the people who write with 'authority' on the subject are wrong.
The one good source of information I've found - and the only one I hand over to artists who encounter this - is the sticky thread about wavy normal maps on this very forum.
Well, while breakdowns and explanations about wavy bakes are indeed useful, the artefacts in the work of the OP aren't caused by that. They are more than likely caused by the baked pixels of the normalmap instantly jumping from one value to another across the (hard) edge but without UV room to breathe since the UVs have been left continuous instead of being split off. Hence the texture filtering causing improper values to show up on both sides. Or perhaps I am misunderstanding it ; either way, showing the map and the UVs overlaid on top as suggested by Celosia would help.
If anything this one case of hard edge artefacts is great for educational purposes as it is quite clear and self-contained. Looking forward to the OP solving it and this thread perhaps becoming a useful learning resource.
Absolutely agree — this is a solid breakdown. The artifact indeed looks like a UV continuity issue rather than a bake smoothing problem, especially when the normal map values shift sharply across a hard edge. Seeing the UV layout over the normal map would clarify it instantly. Automation of manual work – developers can automate repetitive tasks such as reporting, customer communication, or order processing, saving you time and money. That’s why many studios choose to hire game developers when they want to streamline workflows and focus more on the creative side of production.
Here is an simple explanation of vertex normals ChatGPT gave me . Decades ago it took me a month probably of miss and hit to figure out. I wish there were something like chatGPT back then:
[moderator edit, AI summary removed, we don't want AI regurgitated info on Polycount,]
ps. Well , when I asked Chat to generate picture examples it still went not that good although
Am I wrong thinking, if someone wanted to ask ai instead of people, the person would have done so? I feel like filling a forum with chathgpt generated content is just not a great idea.
On the other hand AI chewing on itself, because everything it finds is its own generation will only help deteriorate it...
Everything this thread needs has been answered in the link posted above btw, with practical examples and everything. Why even ask AI in that case?
Not much to add. Neox and pior have already covered most of what's
relevant to the original question about the normal baking artifact,
including the how and why. The handplane video in the thread Neox linked
to is a really good
summary of how it all works. Definitely recommend watching it: there's a
visual breakdown of what typically causes these types of normal
artifacts at the 4 minute mark.
Hiding, rather than deleting, the now slightly off topic portion of the reply:
The removed summary about vertex normals, shading behavior, and
gradation in normal bakes wasn't completely inaccurate but it was
somewhat reductive. Rather than just taking things like that at face
value it can be helpful to look at actual technical documentation that
has visual examples then try applying that knowledge to a few sample
bakes and compare the results. The cube smoothing example it kept
referencing is easy enough to work through and can be used to test
whether or not the points in the summary were accurate.
A couple of other relevant technical resources that have information summaries:
Here's an example of what the cube shading exercise could look like.
(Note that with normal baking it's possible to make the flat shaded cube look like it's smooth shaded and vice versa.)
It's also worth mentioning that normal gradation in bakes is a bit
more nuanced than what's covered by this simple demonstration cube.
Here's a couple of other posts that compare different smoothing split
placement and low poly optimization strategies:
I merely meant Chat could give simple, clear and still deep enough explanation with just a few well structured paragraphs. If asked properly. While many human made instructions just tell : do it this way here or that way there. And thats all . If you don't understand WHY you still know nothing.
I understand why one would use it in general, i also understand why you use it. I dont understand mixing the output that doesnt add anything to this forum here.
I merely meant Chat could give simple, clear and still deep enough explanation with just a few well structured paragraphs. If asked properly.
But, somewhat ironically, in this instance, it could not. Point 1 was solid, but then it fell apart and only gave a reasonable sounding reply without ever answering the actual question (the solution + explanation to which was given multiple times in this thread, according to the available context).
Replies
Also the low-poly seems to have orthogonal but not-connected faces.. and n-gons.. (even if that part seems to lokk okay).
"okay, so every hard edge has to have seams?"
To understand it fully, simply observe the current normalmap of this area (and also post a screenshot of it here, for the sake of future people potentially viewing this thread) :
The cause of these artefacts will be pretty much self-evident once you look at it closely.
Hi,
I am suggesting you to post a *screenshot of the normalmap* corresponding to this area. Once you do so and take the time to observe it closely, you'll understand your issue.
Also, @gnoop , I hate to be a bit of an ass but "Google and try to understand face weighted vertex/split normals or ask chat GPT?" or "something something videocard" isn't contributing anything to help the OP understand the issue. If anything this is just adding even more misinformation/slop to a topic that has been hard to grasp for artists for years.
In case there's no image under the UV you can select it from the header in the UV Editor or click on it from a Shader Editor and it'll appear automatically, but in this case having the normal map itself visible isn't necessary to see how the object was unwrapped; you'll see it highlighted as geometry on the left if you selected the vertices (or use UV Sync Selection).
this is a subject that cannot be googled because at least half the people who write with 'authority' on the subject are wrong.
The one good source of information I've found - and the only one I hand over to artists who encounter this - is the sticky thread about wavy normal maps on this very forum.
https://www.artstation.com/artwork/rlGqrE
If anything this one case of hard edge artefacts is great for educational purposes as it is quite clear and self-contained. Looking forward to the OP solving it and this thread perhaps becoming a useful learning resource.
[moderator edit, AI summary removed, we don't want AI regurgitated info on Polycount,]
ps. Well , when I asked Chat to generate picture examples it still went not that good although
I feel like filling a forum with chathgpt generated content is just not a great idea.
On the other hand AI chewing on itself, because everything it finds is its own generation will only help deteriorate it...
Everything this thread needs has been answered in the link posted above btw, with practical examples and everything. Why even ask AI in that case?
http://wiki.polycount.com/wiki/Normal_Map_Modeling#Smoothing_Groups_.26_Hard_Edges
Hiding, rather than deleting, the now slightly off topic portion of the reply:
A couple of other relevant technical resources that have information summaries:
https://en.wikipedia.org/wiki/Normal_mapping#Calculating_tangent_spaces
Here's an example of what the cube shading exercise could look like.
It's also worth mentioning that normal gradation in bakes is a bit more nuanced than what's covered by this simple demonstration cube. Here's a couple of other posts that compare different smoothing split placement and low poly optimization strategies:
https://polycount.com/discussion/comment/2759935#Comment_2759935
I merely meant Chat could give simple, clear and still deep enough explanation with just a few well structured paragraphs. If asked properly. While many human made instructions just tell : do it this way here or that way there. And thats all . If you don't understand WHY you still know nothing.
I dont understand mixing the output that doesnt add anything to this forum here.