Indeed it’s amazing ! And indeed I completely missed these improvements !
How cool is that ? :
4 seconds !!
Thanks a lot Chris !
Indeed it’s amazing ! And indeed I completely missed these improvements !
How cool is that ? :
He’s definitely one of my new heroes in the recent months. So many important fixes that made Blender UVing a pain to use. He wanted to make a code blog thing if I remember correctly, but I never saw it.
Just last week a new fast packing solution made it in, making it super fast. I did a lot of smart UV layouts for messy photogrammetry/lidar.CT meshes with millions of polys (a thing auto UV in literally every software was struggling with a lot) - now it takes seconds instead of hours. No kidding, sometimes it took Blender a day to finish it, now its half a minute.
It’s definitely worth skimming over his weekly reports - https://wiki.blender.org/wiki/User:Chris_Blackbourn/WeeklyReports
15 years ago, I was known more as a teenager who had an emotional attachment to Blender (especially the game engine) and was among the many who did not know enough about the industry at large to grasp just why professionals did not want to use it (this was before I learned about Blender not having a number of useful operations that Maya and Max users took for granted).
Hence the reason for my quip about the code in what is now the first post. The code had to be doing something that was rather unoptimized such as single threaded loops over every element (which used to be more common throughout Blender). Simple loops can allow a feature to be thrown together a lot more quickly, but there are few ways to make code go slower (without intentionally writing bad code) unless you do the same thing in an interpreted language like Python.
Mayhaps. I couldn’t find the paper about the mentioned “Alpaca” UV packing strategy. To be fair, new algorithms and people coming up with more efficient solutions is a widespread thing, happening to many areas. I don’t know anything about the code there was before, but I’m sure whoever came up with the initial solution was smarter than I’ll ever be.
It’s not like any other software out there could autoUV huge messy meshes like lidars or CTs without any hitches (or at all, in most cases).
I just take issue with how you can casually insult someone who has been working on blender for decades, especially while being completely ignorant of the technical reasoning behind decisions.
Sure, it was single threaded code. 20 years ago was 2003. The Pentium D and Athlon 64 x2 weren’t even released until 2005. Why is that a reason to demean the developer again? Because he isn’t using technology that wasn’t even available yet?
I see you edited your post to be slightly less inflammatory, but please try to be more honest and less emotional in your tone going forward. Improved code is something to celebrate, but you don’t need to be so inflammatory. There is no reason to flame blender or it’s developers to celebrate an improvement.
I can tell you from working on Unwrap Me that it’s not a simple area to work in and kudos to any developer that takes on this beast. You severely underestimate how much effort is involved.
Sometimes when I unwrap something with cylindrical parts I get perfectly straight UVs now in some areas but in other areas I get a C shape. Is there any way to force all the cylindrical areas to automatically unwrap to a straight rectangle?
There’s a really interesting wip from Chris here, about grid unwrap - https://projects.blender.org/blender/blender/issues/78402 This may or may not be what you need, should probably help though.
Definitely something he needs to bring over to the new site, but things like the packing are definitely much bigger issues that are rightfully taking priority. There is also the long-awaited SLIM algorithm which will hopefully be committed before the contract is up.
Yep, that’d be neat. Also this -
" Calendar Week 34
August 30 - September 5
Writing a blog post about recent changes in the UV Editor and underlying code. Currently have outline, first draft of introduction and first section."
They’re probably waiting until some bigger patches made it to master.
Btw you can try the new packer with a test build here:
Man, that is super nice. My current method is using Follow Active Quads after cutting off the parts I don’t want to follow the grid and then sew them together. This is such a huge time saver.
By the way, do we know anything about Chris’ long term plans for UV editing? Would be nice if there was a way to ask him about his scheduling and if it would be possible to give feedback.
One tool that I am really bothered by not being in base Blender is setting texel densities for UVs based on the size of the object in world space and texture map size. Since I am in game development, it is really troublesome when you need to follow very specific guidelines for how many pixels per meter/cm/etc. you should have to keep consistent quality of your assets. Without this tool you need an add-on to do it for you and while there are free ones out there like TexTools, it is still a gaping hole in Blender’s UV mapping capabilities.
Not sure how much of these are relevant right now, but this old draft for dev blog post from last year shows some ideas:
https://wiki.blender.org/wiki/User:Chris_Blackbourn/BlogDraft
Blender Future
And what comes after that? That’s the great thing about blender and open source development, YOU get to decide.
- More game developer features.
- Better UDIM support.
- Support for million+ polygon architectural scenes.
- Automatic seam placement.
- Optimized UV layout based on an existing Lightmap.
- Maybe it’s a One-Click “Just Unwrap It For Me” button.
- Something else?
Get involved and let us know what you want!
Okay, cool! Can’t see any contact info though.
Whenever you contact someone you might point him/her to SureUVW addon for calculating textel density stuff
The Xatlas based packing algorithm was just committed, so will be in the next 3.6 buildbot build!
UV: Add option to Pack UVs using the xatlas strategy · e0d05da826 - blender - Blender Projects
Based on : https://github.com/jpcy/xatlas
He’s around blender.chat’s modelling module channel. It’s probably not a great channel for lengthy feature requests, feedback, or suggestions (primarily for developers after all), but I think it’s okay to simply direct him this thread for feedback.
I hope that not only UV editing will improve, but also the mesh editing and curve editing mode.