Image¶
Loads an image file from the configured input directory (or via upload) and outputs a normalized image along with an optional mask. Supports multi-frame formats (e.g., animated GIFs) by batching frames when sizes match and generates a mask from the image’s alpha channel (inverted, so 1 = masked area).

Usage¶
Use this node at the start of a workflow to bring an external image into the pipeline. Commonly paired with encoders or preprocessors. Select a file from the input directory (or upload via the UI control) to get an IMAGE output (RGB, normalized 0–1) and a corresponding MASK output derived from transparency.
Inputs¶
| Field | Required | Type | Description | Example |
|---|---|---|---|---|
| image | True | filename (selectable/uploadable) | The image file to load from the input directory. The UI shows a list of available files and also supports uploading a new image. | example.png |
Outputs¶
| Field | Type | Description | Example |
|---|---|---|---|
| IMAGE | IMAGE | The loaded RGB image, normalized to 0–1. For multi-frame images with consistent dimensions, outputs a batch of frames. | A normalized image tensor suitable for downstream image processing nodes. |
| MASK | MASK | A mask derived from the image’s alpha channel, inverted so that 1 indicates masked (transparent) regions and 0 indicates unmasked (opaque) areas. If the source has no alpha, outputs an empty mask. | A mask aligned with the IMAGE output; for single-frame images, a single mask; for multi-frame, one mask per frame. |
Important Notes¶
- Input location: Files are read from the platform’s designated input directory; the widget also supports image upload.
- Color and orientation: Images are converted to RGB and auto-rotated according to EXIF data.
- Normalization: Pixel values are normalized to a 0–1 range.
- Alpha handling: If an alpha channel exists, the mask is generated as inverted alpha (1 = transparent/masked). If no alpha, an empty mask is produced.
- Multi-frame support: Multi-frame images (e.g., GIF) are batched only if all frames share the same dimensions; frames with mismatched sizes are skipped. MPO multi-frame batching is excluded.
- Consistency requirement: When batching multi-frame images, all included frames must have identical width and height.
Troubleshooting¶
- Selected file not listed: Place the image in the input directory or use the upload control, then refresh the file list.
- Unexpected empty mask: The source image likely has no alpha channel. Provide an image with transparency or generate a mask using a separate node.
- Multi-frame image outputs only one frame: The format may be excluded from batching (e.g., MPO) or frames may have differing dimensions; ensure consistent frame sizes and a supported format.
- Orientation looks wrong: Verify the source image’s EXIF rotation. The node auto-applies EXIF transpose; if results seem off, re-export the image without problematic EXIF data.
- Color/brightness appears different: The node converts to RGB and normalizes to 0–1; downstream nodes must expect normalized RGB input.
- Load error or corrupted output: Ensure the file is a valid image and not corrupted; try re-saving the image in a standard format (PNG/JPG).
Example Pipelines¶