Huge Edit: I have moved to a NUKE and Modo workflow for the feature "Puddles of Light", focusing on deep compositing. Has changed most of my workflow as previoiusly stated (which was AE/Blender). I still use AE and blender a lot, but I couldn't get the seamless integration out of them with out a ton of tweaking. I will start a deep compositing low budget thread perty soon at pudl.tv
There's not much written on 4k vfx pipelines with after effect and blender, so I could use some community help/experience with this. I have been doing tons of tests, back and forth with idea's, and finally came up with this work flow. This work flow will be locked in the next couple of weeks. [edit ... film spec. is now locked for "Puddles of Light" ]
"Puddles of Light" Technical Specification
Starting point is r3d 4096x2160 24 fps at 180 degree shutter ussing RPP with T1.8 depth of field. Secondary Motion controlled shots at T8 (for 2.5D calculations). Shots are motion controlled with laser measured rooms. Shots are framed at a 2.39 aspect ratio (4096x1716), with 3d markers placed in the cropped regions from the full container.
Camera Color Configuration:
Filters - IR Tiffen Mirror Filter is on at all times, ND Tiffen as needed, no other filters.
ISO - 320
REDCODE - 7:1
Kelvin - 5k (no matter keep that setting)
All shots start with a RED Color chart and focus chart, most scenes are not time coded.
Directory structure is organized by scenes (1-36), with upto 5 vfx sections per scene.
Step 1 - In adobe/redcine do a basic grade (iso,redlogfilm, log adjust) - no fine detail color grading, try to have no digital grain
Step 2 - output from adobe at 4096x2160 to openexr and png, and 2k png
Step 3 - In blender use openexr camera solutions and textures, and 2k png for background alignment (i.e. when doing animation I reference from the png but all the textures manipulated are from the openexr).
Step 4 - flat light render from blender to openexr 4096x1716 with alpha channel (70% of the area is ussually alpha) [zdepth was in openexr, now doing in png ... shadow mask is also in png]
Step 5 - build feather masks in AE for blender 4096x1716, perform depth of field blur based on zdepth
Step 6 - composite the r3d at 4096x1716 with the feather mask from step 5
Step 6a - inside AE use Color Finesse to align the VFX with the R3d footage during compositing
--- repeat this for all sections of a scene
Step 7 - pull in ae comp's to premier, only cut and transitions in premier (no grading), output in dpx entire scene (i have 36 scenes btw)
Step 8 - pull all dpx files from all the scenes into davinci, add blur and grain with final color grades
Step 9 - output from davinci in dpx
Step 10 - read in step 9 dpx into adobe, for sound scoring in 2k
Step 11 - dcp programs for 4k & 2k