Skip to content
nitida

TransformOptions

type TransformOptions = {
  dpr?: 1 | 2 | 3;
  duration?: number;
  effect?: TransformEffect;
  fit?: TransformFit;
  format?: TransformFormat;
  gravity?: TransformGravity;
  height?: number;
  quality?: "auto" | number;
  start?: number;
  width?: TransformWidth;
};

Defined in: packages/asset-client/src/transform.ts:67

Properties

Section titled “Properties”

dpr?

Section titled “dpr?”
optional dpr?: 1 | 2 | 3;

Defined in: packages/asset-client/src/transform.ts:87

Device pixel ratio. Width/height are multiplied by this before resize.

duration?

Section titled “duration?”
optional duration?: number;

Defined in: packages/asset-client/src/transform.ts:116

Video-only: clip duration in seconds (1..300). Image transforms ignore. With start, lets a single request grab a sub-clip.

effect?

Section titled “effect?”
optional effect?: TransformEffect;

Defined in: packages/asset-client/src/transform.ts:106

AI effect applied before resize/encode.

  • removebg: remove the background; output is a transparent PNG of the foreground subject. Forces format=png regardless of other format hints. Runs U²-Net ONNX locally (or BRIA via Replicate when BG_REMOVAL_BACKEND=replicate). Single cache miss per (sha, dsl) tuple; subsequent identical DSLs serve from R2 — no inference, no per-image cost.

  • genfill: aspect-extension outpaint via Flux-Fill Pro on Replicate. Requires BOTH width and height — the server fits the source centered into the target canvas and outpaints the gutters. Output is PNG (forced) at exactly target dims. ~$0.05/image first time; same R2 cache as removebg after. Primary use case: building OG cards (1200×630) from portrait listing photos without awkward edge mirroring.

fit?

Section titled “fit?”
optional fit?: TransformFit;

Defined in: packages/asset-client/src/transform.ts:79

Resize fit mode. Default cover server-side.

format?

Section titled “format?”
optional format?: TransformFormat;

Defined in: packages/asset-client/src/transform.ts:83

Output format. auto → policy decides (see asset-manager bench).

gravity?

Section titled “gravity?”
optional gravity?: TransformGravity;

Defined in: packages/asset-client/src/transform.ts:81

Crop gravity. auto uses sharp’s attention strategy.

height?

Section titled “height?”
optional height?: number;

Defined in: packages/asset-client/src/transform.ts:77

Target max-side height. Multiplied by dpr server-side.

quality?

Section titled “quality?”
optional quality?: "auto" | number;

Defined in: packages/asset-client/src/transform.ts:85

Output quality. auto → format-specific default.

start?

Section titled “start?”
optional start?: number;

Defined in: packages/asset-client/src/transform.ts:111

Video-only: clip start in seconds. Image transforms ignore. Accepts decimals (e.g. 1.5 for sub-second seek).

width?

Section titled “width?”
optional width?: TransformWidth;

Defined in: packages/asset-client/src/transform.ts:75

Target max-side width in CSS pixels (multiplied by dpr server-side). MUST be a TRANSFORM_WIDTHS value — off-ladder widths are rejected (HTTP 400) by the edge whitelist for unsigned URLs, so the type forbids them at compile time. For SIGNED URLs with a custom width, use SignedTransformOptions (which widens this to number).