Recipes

Hand-written helpers for common jobs, built from the same generated ops. Each key-to-key recipe returns a ready-to-submit Graph.

@jtdigital/renderbox-sdk on npm →

faceRedact

faceRedact(input: string, output: string, opts?: FaceRedactOpts): Graph

Detect faces, track them, and redact: the GDPR pipeline in one call. Equivalent to open, detect, track, redact, write. Inputs/outputs are S3 keys.

Param	Type	Description
model	string	Face detection model (default 'retinaface_mv2', permissive: Apache-2.0 code + MIT weights).
threshold	number	Detection confidence threshold (default 0.5).
tracking	'bytetrack' \| 'sort' \| 'deepsort' \| false	Tracker for stable per-face identity; false redacts raw detections.
redactMode	'blur' \| 'pixelate' \| 'black'	Redaction style (default 'black').
redactRadius	number	Blur radius / pixel block size.
output	SinkAvOpts	Encoder options for the output file.

redactPasses

redactPasses(video: VideoStream, passes: readonly RedactPass[]): VideoStream

Burn any number of detector passes onto one video, sequenced as data — one { detect, track?, redact? } per pass. Every pass detects on the ORIGINAL frames (an earlier burn must not hide what a later detector needs to find) while the burn accumulates. Defaults: ByteTrack tracking, { mode: 'black' }.

Param	Type	Description
detect required	DetectOpts	Detector options for this pass (model, threshold, stride, batchSize, …).
track	TrackOpts \| 'bytetrack' \| 'sort' \| 'deepsort' \| false	Tracker — an algorithm name or full TrackOpts (e.g. persistFrames); false redacts raw per-frame detections (default 'bytetrack').
redact	RedactOpts	Treatment for this pass — variant-typed: blur requires radius, black forbids it (default { mode: 'black' }).

depthBlurPipeline

depthBlurPipeline(input: string, output: string, opts?: DepthBlurPipelineOpts): Graph

Estimate monocular depth and apply distance-aware background blur: portrait mode for video.

Param	Type	Description
depthModel	string	Depth estimation model.
maxBlur	number	Blur strength at the far plane.
nearPlane	number	Depth kept fully sharp.
farPlane	number	Depth fully blurred.
output	SinkAvOpts	Encoder options.

makeSlideshow

makeSlideshow(output: string, opts: MakeSlideshowOpts): Graph

Build a slideshow from stills: per-slide Ken Burns motion (12 presets, from gentleZoomIn to parallax), crossfade transitions, and an optional music bed with fade in/out.

Param	Type	Description
slides required	SlideConfig[]	Slides: { path, duration, kenBurns? } (paths are S3 keys).
resolution required	[number, number]	Output resolution, e.g. [1920, 1080].
fps	number	Output frame rate.
transition	SlideshowTransition	Transition type + duration between slides.
music	SlideshowMusicOpts	Music bed: { path, volume?, fadeIn?, fadeOut? }.
output	SinkAvOpts	Encoder options.

letterbox

letterbox(input: VideoStream, w: number, h: number, opts?): VideoStream

Fit a stream into a target frame, preserving aspect with padding. A stream transform, so it composes inside a larger graph.

snip

snip(input: VideoStream, start: number, end: number): VideoStream

Frame-accurate clip extraction with timestamps reset to zero.