Thanks to the contribution we get in 2014 from the EU Media Fund we will involve a large team of developers for the film project. Read more about this at the code blog.
Right after the script workshop, I invited a group of developers for a technical workshop day. Aim was to inform them on the status of the project sofar, and help mapping out the topics for the pipeline and development in general.
- Bart Veldhuizen – works part time as Gooseberry ‘quarter master’ to investigate which asset/content/project management system we should choose and/or develop.
- Francesco Siddi – full time production coordinator. Brings in experience with project/render managing tools and as artist.
- Brecht van Lommel – lead Cycles developer. Works for Blender Institute.
- Lukas Toenne – currently finishes work for Development Fund, will work full time for Blender Institute on Gooseberry topics.
- Antonis Ryakiotakis – starts working per March for Blender Institute on Gooseberry topics.
- Jeroen Bakker, Monique Dewanchand – from Atmind, the developers of Blender’s compositor.
- Ton Roosendaal – that’s me, in role of producer Blender Institute.
A whole lot of topics were reviewed and discussed. Find a summary below:
ASSET / VERSIONS / PROJECT SYSTEM
The investigations we will do for asset/project management is going to be a separate discussion and blog post here. Here’s some notes:
- Currently investigations include TACTIC (from SouthPawTech).
- Director requested a great ‘story editing’ tool – which could be web based, so everyone can see the film in every quality and version at any time. Idea: this could be main interface to access assets and projects.
- Better define what part of an asset is handled through an asset manager. Coordinate with Andrea Weikert, see which parts of asset browser branch can be refined to work for Gooseberry asset requirements.
- Need to define design of asset sharing system. Where is it a new tool, where is it in Blender itself.
- Investigate ‘compiling’ of .blend files, to pack in dependencies only, and allow server-side unpack before versioning happens.
We wait for further investigations and feedback/interviews of people before deciding or defining work. Comes back in March!
UNIFIED PHYSICS SYSTEM
The animation system and physics system should work together much better… loads of hair and fur and cloth is going to be needed for the film! Here’s some topics:
- Better and more generic handling of cache file storage, integrate into asset management. Investigate Alembic for it.
- Layered animation/simulation:
– Allow simulation data as input for other simulations (“depsgraph” for simulations), e.g. hair on cloth, particles on top of fluid sims.
– Caching (baking) should turn simulations into animation data
– Tools then can work on such cached data too (sculpt, hair edit)
- More unified basic abstract models for simulation (timesteps, cache input/output, solver functions, friction, force units & fields).
- Investigate Bullet for softbody replacement. How’s Bullet cloth coming?
- Investigate papers and existing libraries for modernizing hair sim.
- Unify collision handling, use Bullet collision detection algorithms where feasible
Antonis will further intestigate hair and cloth simulation.
Lukas will continue work on a new node based Particle system. This system will be separated from a hair/fur system, which also will need an upgrade or partial recode.
- Two main features missing: volume rendering and deformation motion blur.
- Besides that, focus on CPU rendering optimization. Especially hair rendering and complex scenes in general.
Brecht is going to lead this work.
- Blender Internal and Freestyle will be used for NPR rendering.
- Freestyle performance needs to improve, can take a long time before you see any results.
- Get better NPR shading features.
Ton will connect with Tamito and Shinsuke about scheduling this work.
- Finish and integrate GSOC 2013 tools
- Handle pending requests from community (multires performance, dyntopo undo)
- Improve hair painting/grooming tools.
Will be handled by Antonis.
The system is falling behind in development a bit, we need to get more developers active here too. Lukas and others expressed concerns that the current code is too hard to grasp, and in need of some cleanup.
Agreed was to have a session here in Blender Institute with the key developers to dig in the code structure together, make sure everyone understands how it’s (meant to) working, and then decide on architecture fixes or not.
Next then will be prioritizing work – on usability, interaction and tools, but also on ‘canvas’, percentage (preview) renders, shot caches, etc.