# Image Resize

One resize node for the jobs that usually pull in a whole pack: hit an
exact **width+height**, scale until the **longest** or **shortest edge**
lands on a number, meet a **megapixel** budget, or multiply by a **scale
factor**. An optional mask rides through the identical geometry, and the
output width/height come back as INTs ready to wire onward.

## Target modes

**Target** (`target_mode`) picks the mode. The card shows only that mode's
size row, and the other size values are ignored.

- **Width × height** (`width+height`): The two **Size** fields, W and H
  (`width`, `height`). Their input sockets sit with the node's inputs,
  under `image` and `mask`; a linked one greys its field out. `0` keeps that
  source dimension, and with only one set the other follows the source
  aspect.
- **longest_edge** / **shortest_edge**: Scales until that edge equals
  **Edge** (`edge_length`) exactly; the other edge keeps the aspect. `0`
  keeps the source size.
- **megapixels**: Scales (aspect preserved) until width × height is about
  **Budget** (`megapixels`) million pixels. `0` keeps the source size.
- **scale_factor**: Multiplies both dimensions by **Scale** — `0.5` halves,
  `2.0` doubles. `0` keeps the source size.

## Fit

**Fit** (`keep_proportion`) decides what happens when the target aspect
differs from the source:

- **stretch**: Distorts straight to the target.
- **fit**: Shrinks the target box to the source aspect — the output can be
  smaller than requested, but there are never bars and nothing is cropped.
- **cover** (`cover_crop`): Fills the target completely and center-crops
  the overflow.
- **pad**: Fits inside the target and fills the rest with **Fill**
  (`fill_color`: hex, `R, G, B`, one grayscale number, or a CSS name). The
  new bars are `1.0` in the mask output — the same generated-area contract
  as the pack's other pad nodes, so it feeds an inpainter directly.

## Notes

- **Multiple** (`divisible_by`) snaps each output dimension to the nearest
  multiple (never below one step) — `16` for WAN, `8` for most latent
  spaces. It wins over exact proportion, so a dimension can shift by up to
  half a step. It applies even when everything else says "keep the source".
- **Pixels are only ever invented on explicit request.** In every target
  mode except `width+height` the box is derived from the source's own
  aspect, so there is nothing to letterbox against: the proportion modes
  all resolve a `divisible_by` snap with an invisible sub-half-step
  resize, and the mask output stays black. Bars — and white in the mask —
  can only appear when you set a `width+height` box that disagrees with
  the source and choose `pad`.
- **Filter** (`interpolation`): `lanczos` (PIL, in float — the sharpest
  all-rounder), `bicubic`, `bilinear`, `nearest` (pixel art, hard masks),
  `area` (best for strong downscales). A resize that changes nothing passes
  the tensor through bit-identical.
- The mask output is the input mask through the same transform, plus the
  pad bars; all zeros when no mask is wired. IMAGE is BHWC, MASK is BHW,
  and batches flow through frame by frame.