Video codec; h264 through libx264 is the only encoder path.
libx264 CRF quality factor, passed to ffmpeg unchecked. The caller supplies a supported encoder value; this field alone does not certify perceptual quality or repeatable encoded bytes across encoder versions.
Shared output clock and pixel geometry. Take the caption and pose-keypoint
plans' fps, aspect and crop from this same object so every companion is
sampled against the render it describes.
Pixel format passed to the encoder; the supported path uses yuv420p.
Id of the shot or sequence to render.
Tone mapping applied before the sRGB encode: none for the stylized toon
path, acesFilmic for the photoreal path.
This is the delivery default, and a scene overrides it. A scene's optional
environment owns the photographic response of that scene (its image
lighting, exposure and curve), and one render spec covers a whole sequence
of scenes, so a spec-level curve forced onto every shot would flatten a
night interior and a noon exterior into one response. A scene that
declares no environment keeps this value.
Structural passes (depth, normal, mask, outline) ignore both. Their pixels are geometric facts, so tone mapping and exposure are bypassed there.
Render parameters for turning a shot or sequence into frames and video. Kept separate from the shot so the same shot renders at a draft 12 fps or a final 30 or 60 fps without editing the content.
The pipeline:
N = round(duration * frameFormat.fps)frames, sampled att = i / frameFormat.fps(rational, not accumulated, for determinism), each rendered headless to an sRGB image, then encoded byffmpeg -framerate fps -c:v libx264 -pix_fmt yuv420p -crf crf -r fps -s widthxheight -movflags +faststart. The encoder never tone maps; the curve is a renderer setting applied before capture.