FASTER EVERYTHING | Blender.Today LIVE #162

Back with dev news!

BLENDER TODAY #162

NIGHTLY BUILDS NOW ON STEAM AND SNAP

ASSETS
PERFORMANCE

RIGGING

GEOMETRY NODES

UI/UX

Read all updates:

--

Pablo Vazquez Links:

#BlenderToday #b3d

[DirkTeucher]

4:17 - Steam experimental branch
6:37 - Assets - Add an active library per workspace
7:00 - Assets - Initial asset list as part of asset system design
9:00 - Performance - Multi threading auto smooth sharp edge calculation (10 x faster)
10:15 - Improve speed of Delaunay when there is a lot of holes (72 times faster)
11:35 - Speed up Delaunay raycast (2 x faster)
12:30 - Make it optional to track input> output mapping in delaunay_2d_calc
13:00 - Edit mesh editing vertices (1.4 x faster)
13:53 - Draw cache extract tris in parallel ranges (8% faster )
15:00 - Optimize edit-mode undo (11% speedup with 1.5 million polys)
17:06 - Compositor: Full frame render layers node (2.5 x faster)
17:20 - Compositor: Full frame ellipse mask node (3 x faster)
17:20 - Compositor: Full frame brightness node (2.3 x faster)
17:30 - Compositor: Full frame mix node (2.3 x faster)
17:30 - Compositor: Full frame Movie clip node (2 x faster)
18:20 - Speedup rigid body "Copy from active" operator (200 times faster)
19:50 - Surface deform: Support sparse binding mode for improved performance (faster, less disk space less memory)
23:20 - Geometry nodes - Socket inspection
24:26 - Geometry nodes - Display node warnings in modifier
25:26 - Geometry nodes - Curve primitive Quadrilateral
27:00 - Geometry nodes - dim links whose start and end sockets are not visible
28:35 - UI/Assets - Initial asset view UI template
31:50 - UI: CTRL + F to search inside UI list side panel
32:30 - UI: Support left right arrow key walk navigation in UI lists
33:00 - UI: support UI list tooltips, difined via Python scripts
33:21 - UI: Internal support for custom UI list item drag and active operators
34:15 - UI: Auto scroll to keep active text buttons in view
35:30 - Papercut: Flip driver debug UI
36:18 - Keymap: use Alt-Q instead of D for transfer mode operator
36:48 - Keymap: use Shift-Tab to toggle snap in the sequencer
37:30 - VSE: Speed effect layout update
38:30 - XR: Reference space improvements
40:15 - Weld modifier: add "loose edges" option (better cloth modelling)
40:58 - Fix: Subtitles: Order of channels in the .blend file will be kept in the .rst
41:45 - Fix: Allow keyframing vertex mass in cloth sim
42:10 - VSE: Make pasted strip active
43:30 - Refactor: Move vertex group names to object data (PSA 3.0 save do not open 2.93 or data lost)
44:21 - GPencil Tools: Box deform multi object edit
44:35 - Rigify: Allow adding a custom root bone to the metarig
44:45 - Rigify: Add utilities for working with rig instance parent chains
45:00 - PyGPU: New method 'GPUFrameBuffer.read_depth' (Addon makers)
45:50 - CyclesX - Cleanup remove disabled OpenCL implementation
46:21 - CyclesX - Make Optix 7.3 minimum required SDK version
46:25 - CyclesX - Support shadow catcher behind transparent object
47:40 - AMD drivers for CyclesX coming soon

49:27 - Questions

[KuruGDI]

22:00 Ohhh! I wanted to have this _show value_ thing for a really long time since sometimes it was hard for me to analyse what went wrong when I tested around in geometry nodes. This will be really helpfull!

[barnabasch]

I love the updates, and massive respect to the team! One thing I find kinda very important to add would be viewport/camera changes undo. Cheers!

[petebateman]

My threadripper is finally getting some blender love!

[lucashaymaker]

Pablo Vazquez: "you can make mistakes faster"
🤣🤣🤣

[julinaut]

20:53 everyone who got that reference is probably really cool ^^

[noelmacaoidh]

Whole videos on specific updates would be very welcome ... thank you ...

[qwertasd]

With all the procedural nodes we slowly get into procedural descriptive drawing ie a cube with 3mm rounding. Can this bring us closer towards CAD drawing standards eagis etc?

[jendabekCZ]

Created 500k mesh, tried to Weight and Vertex Paint, both still slower than in 2.79, so will keep waiting ...

[tomkayak]

Yes I enjoy an episodes devoted... leaving now to watch it. Be back soon :)

[philhacker]

Blender 3... Performance and multithreading. Overall more information. Geometry More Nodes.
A section on multithreading in the manual would be great. So more can develop multithreading add-ons.

[KuruGDI]

We need more blender related T-Shirts like there was for the 2.8 codequest!

[reutofir]

I didn’t know about Blender until I watched @kutiman ‘s video called Guruji -amazing! 🔥🔥

[DirkTeucher]

01:00:00 - Collections vs groups - Yes I agree with Pablo except I don't care if it happens by adding groups or sticking with collections, I am happy either way I just would like to see:

1) Selecting a collection selects all the objects in that collection
2) When a collection is selected allow it to be transformed and keyframed to transform all the objects together around the collection origin
3) Have an option/toggle to toggle the transform of the objects individual origins when say the collection is rotated or scaled. This is awesome for mograph stuff if you can nest collections inside the parent and then rotate the top level parent to let the children rotate in different ways.
4) I don't care if adding groups is a better solution according to the developer. I just want 1, 2, 3 and having another layer or not should be up to the developers to figure out what is best in the code architecture base.

[holandonhdunto]

Well there could be a STOPMOTION mode like the 2d blender mode, or good STOPMOTION and you can make sketches and a base to draw lines on, for example 2d characters could be merged with 3d objects with STOPMOTION, that way it would give to make complex animations such as mouth movements, eye blinks, clothes movement without using physics, and with the advantage of using the sculptor mode, you can use freestyle and so on. In the same way that you can animate lines in 2d mode, meshes can animate replacing the shaper keys, and using frame modifiers similar to 2d.

[JacobMerrill]

Occlusion Culling, LOD, and GPU skinning next please.

[Paruthi.]

Yes please..
Developer updates will be more informative and interesting

[ramanmono]

Lol, AMD experimental build when 3.0 already out. So they gonna rush it and it will be the same as before. I bet RX 6800 XT gonna be more than 3 times slower than RTX 3070. And obviously the ray accelerators are gonna be left unused.
As I understand it, now some library is gonna translate the cuda code to HIP C++ code which should be able to work with new AMD drivers (lets hope there). I would like to know what other program does this successfully for programs not having opencl.
The better way would have been to write the new cyclesX kernel code in HIP C++ directly. Although I have no idea if any other program does this successfully to get proper performance from both AMD and Nvidia GPU's.

[blenderhead]

I make mistakes super fast already, but this is good 🤣

[peterpui]

Nice blue blender t shirt, Pablo! I would like to have Suzanne Monkey on the Blender t shirt on the next?