Skip to content

armgit / vork

Public Imported from GitHub
0

Headless 3D and VFX engine for agents. Research preview.

2.3 KB
Raw Blame

OUTPUT ENCODINGS .png RGBA still (keeps alpha; use render.background "transparent"); t=0 unless --at; other channels beside it as NAME..png PNG filename frame patterns (two or more # characters or printf-style %d/%04d) are usage errors; use a directory or %06d.exr for frames. Parent directories and other filename characters are literal. directory (no extension or trailing /): NNNNNN.png frames plus channel PNGs PATH/%06d.exr: one multilayer OpenEXR per frame, scene-linear premultiplied, before exposure .mp4 H.264; .mov ProRes 4444 RGBA (keeps alpha; transparency needs render.background "transparent") --at SECONDS: one still; needs .png, directory or EXR. --frame-start/--frame-end N: inclusive absolute frame indices, default the whole clip; not with --at. Sidecars: OUT/meta.json for directories; OUT.meta.json for files, e.g. check.png.meta.json (fps, per-frame camera and depth unit, object_ids pass_index -> node, exposure, warnings [{code,requested,dropped}] for omitted/substituted effects); camera.json in directories or OUT.camera.json for files when camera is a channel (videos always; vork help camera-json). PNG channels: depth and object_index grayscale16 (depth metres per unit in meta.json frames[].depth_unit_m); view_normal RGB8 = (v+1)/2; motion RGB16, RG = 0.5+0.5*offset in image extents, clamped to ±1; shadow_catcher RGB16 linear, clamped 0..1; aov linear 16-bit, clamped 0..1. EXR channels: R,G,B,A; Z (metres, background 1e10); view_normal.X/Y/Z; object_index.X; shadow_catcher.R/G/B (unclamped multiplicative light ratio); motion.X/Y (pixels, +Y down, current minus previous); aov .X or .R/G/B; float32. depth: view-axis metres to the nearest surface, not blended across edges. --exr-codec none|rle|zips|zip|piz|pxr24|b44|b44a|dwaa|dwab (default zip; pxr24/dwa are lossy). --diagnostics: OUT/timings.json for directories; OUT.timings.json for files.

shadow_catcher follows Cycles: catcher material light with CG / without CG, including shadows, bounce and reflections. A zero baseline component is neutral 1 without division; light added by CG to an otherwise unlit catcher is not representable. Combined always treats a catcher as transparent black on transparent film; opaque film shows background times ratio. Holdouts are transparent black on transparent film, opaque black on opaque film. Channel selection never changes combined.