I will try to upload the blend file when I get back from work,file is pretty heavy because chracter is animated through shapekeys on all frames (200mb).Topology seems very nice.
I will upload and identify the frames you should look into.
ok,here it is.
There are gaps on;
Frame 210 left ,on lower leg
Frame 213 lef,t on the thigh
Frame 216 right, on leg
Frame 221 right,on arm sleeve
What is interesting is if I change topology a little using the triangulate modifier the gaps disappear BUT other different gaps appear on other frames (also always vertical)
For me, crease threshold and crease fade are giving unpredictable result.
Same settings are giving different results according to mode.
I have no crease if crease fade value is equal to 0. I have more probability to obtain crease edges by giving a negative value to crease threshold. And crease threshold value affects crease fade.
I did not succeed to obtain in between results on sphere. I am stuck to theses 2 choices. No crease edge or all visible edges are crease edges.
Here is the .blend file giving this result with absurd values.
Thanks, I’ll check it when I got time!
That’s probably my last modification for shader broke something stupid, I got this too, will fix it when I got time.
Oh hi! didn’t Know you were here too, I saw the updates and I’ll compile it when I got back.
Hi Christian, I’ve fixed the vertical line gap bug, it is caused by inconsistant triangle cutting threshold value in different places of calculation code. The locations in Frame 210, 213, 216, 221 are all fixed now. I didn’t test more frames, but by principle, this fix should work on all the issues related to this type of the problem.
There are still some imperfections on some parts of the vector calculation code. Please tell me if more line bugs appear. A more accurate and pricision-tolerant improvement is at hand but it will cause some performance drop. So if the condition is okay, then we can postpone this accurancy improvement further.
And… Thanks for the test!
And @zeauro This problem is fixed now. If it still exists that must be a videocard issue, and please send me test file.
And DPIX now also have normal controlled line style. Same configure as software mode.
Download the testbuild:
Yes. it is fixed. Software and DPIX are giving same result for shared supported settings, now.
You might want / need to add some settings for how high you want the pricision to be.
For offline final render you want it to be as high as possible, but for real-time preview a lower setting might be helpfull for those who don’t have a fast system.
Works great!!! no gaps whatsover during the whole animation, I think it could be used in production of japanese animation for example.Just for rendering of course 2.8 is not production ready.
There are some little gaps in self-intersections but I think they are normal,they don’t make the animation flicker.
I will test further with other files. I think you can postpone accuracy improvement…as a renderer lanpr feels pretty robust in my opinion for the moment.
Btw Edge mark does not work anymore,alembic import also disappear there was probably some reverts in the 2.8 branch…
Anyway,big thanx for staying a bit longer beyond gsoc even with school n’ stuff.
EDIt:ok, I built the build myself and now I got Edge mark and alembic working again
Yeah, I could give it a go
I am getting a lot of depth sorting issues when using DPIX, resulting in things popping through. This gets better when i get closer with the camera, but it still an issue.
I had this too until someone suggested I fiddle with “clip start” value in camera panel,I brought it to 35cm and it solved the issue for me(of course your scene might require another value).
Yes, thank you, that did the trick.
How can I animate lines in time to grow it as branches ?
Any help is very welcome.
By compositing obviously.
Thank you for your excellent work and continued development for this new NPR line render engine.
Is it possible to share the .blend file for blueprint, please?
“Excuse me for my bad english , I’am french”
This blend file is from BlenderWiki, which is originally used as freestyle’test file.
Here’s the link https://en.blender.org/index.php/File:M-130Blueprint.zip_the_.blend_file