202310.md 3.9 KB
[LAMDWIKIPOST 20231007181935; VER 20231007181935; ]
unique challenges in the implementation of line art
- cuts the line instead of keeping occlusion result in ambigious state, conforms better to shapes but needs a higher precision of floating point intermediate storage, and not easy have this part of data exposed reasonably to the outside for custom manipulation
- doesn't do fully tiled processing means it's kinda slow sometimes.
- intersection filtering is a headache since it's global.
- the fact that we use GP to rasterize means currently chaining plays a very big role in line quality which posts a design optimization requirement for GP which will talk later.
- Shadow project means doing visibility for 2 times which is slow.
- Contour means doing flag cutting for yet another time, and also very hard to filter under current setup.
[LAMDWIKIPOST 20231007181950; VER 20231007181950; ]
node based line filtering possible design choices
- feed multiple attributes [mockup todo]
- feed line styles [mockup todo]
- possible integration with loop/simulation node, which can solve multiple line type issue [mockup todo] due to multi step projection
- compatibility/versioning with previous modifier setup
- stroke baking design?
- z-reorder: multiple passes [mockup todo]
- need compositor involvement
[LAMDWIKIPOST 20231007181950; VER 20231008052649; HASI 20211224061904.jpg; ]
node based line filtering possible design choices
- feed multiple attributes [mockup todo]
- feed line styles [mockup todo]
- possible integration with loop/simulation node, which can solve multiple line type issue [mockup todo] due to multi step projection
- compatibility/versioning with previous modifier setup
- stroke baking design?
- z-reorder: multiple passes [mockup todo]
- need compositor involvement
[LAMDWIKIPOST 20231007181950; VER 20231008054401; HASI 20211224061904.jpg; ]
node based line filtering possible design choices
- feed multiple attributes [mockup todo]
- feed line styles [mockup todo]
- possible integration with loop/simulation node, which can solve multiple line type issue [mockup todo] due to multi step projection
- compatibility/versioning with previous modifier setup
- stroke baking design?
- z-reorder: multiple passes [mockup todo]
- need compositor involvement
[LAMDWIKIPOST 20231007181950; VER 20231008054547; HASI 20211224061904.jpg; ]
node based line filtering possible design choices
- feed multiple attributes [mockup todo]
- feed line styles [mockup todo]
- possible integration with loop/simulation node, which can solve multiple line type issue [mockup todo] due to multi step projection
- compatibility/versioning with previous modifier setup
- stroke baking design?
- z-reorder: multiple passes [mockup todo]
- need compositor involvement
[LAMDWIKIPOST 20231007181950; VER 20231008054604; HASI 20211224061904.jpg; ]
node based line filtering possible design choices
- feed multiple attributes [mockup todo]
- feed line styles [mockup todo]
- possible integration with loop/simulation node, which can solve multiple line type issue [mockup todo] due to multi step projection
- compatibility/versioning with previous modifier setup
- stroke baking design?
- z-reorder: multiple passes [mockup todo]
- need compositor involvement
[LAMDWIKIPOST 20231007182013; VER 20231007182013; ]
benifits
- doesn't have evaluation conflict in depsgraph
- much better intersection filtering effect
- line styles can be processed further with geometry generation downstream