(Deno) Multi Image Loader
A batch image loader that doesn't turn into a 3000-pixel-tall node
- multi_output
- width
- height
The default ComfyUI Load Image takes one image and gives you one image. Fine for stills, useless for the batch-guide workflows that video models actually want - LTX, Wan, and the multi-reference models all perform better when you can feed a whole set of reference frames in one go. (Deno) Multi Image Loader is the minor-upgrade answer: a batch loader with a fixed-height, scrollable gallery instead of a node that grows into your canvas like a skyscraper.
What makes it worth installing
The core workflow idea is borrowed from WhatDreamsCost's batch-guide approach, then reshaped for daily use. You get a gallery where you can drag to reorder (with a stable placeholder, so you're not fighting the list), upload via button, drag-and-drop, or paste, and an Input Folder browser that reads your existing ComfyUI/input images complete with subfolder navigation - folder tiles, double-click to descend, a Parent button to climb back, and nested images that keep their subfolder paths. Files sort newest-first by modified time, which is a small thing that saves a lot of hunting.
Then the resize handling. The loader resizes everything to one uniform batch size before it hands you multi_output, and it gives you the same size controls as the rest of the Deno pack:
mode-Keep Input Ratio,Preset Ratio, orManual. Default keeps whatever ratio your images already have.megapixels(default 1),divisible_by(default 32),width/heightfor manual work.interpolation- lanczos default, the right quality choice.resize_method-Center Crop (Fill)orFit(letterbox/pillarbox).
Outputs are multi_output (the uniform IMAGE batch), plus width and height so downstream sizing nodes don't have to guess.
How it fits a workflow
This is the front door for LTX-style multi-image work. Wire multi_output into something like (Deno) LTX Sequencer - which even auto-fills its num_images from this loader when connected - and you've got a clean path from "a stack of reference images" to "a sequence with guide keyframes inserted." Same story for Wan 2.2's reference-based workflows, where the loader's reorder-and-paste flow makes iterating on reference order genuinely fast.
Install is the standard pack route:
cd ComfyUI/custom_nodes
git clone https://github.com/Deno2026/comfyui-deno-custom-nodes.git
# restart ComfyUI
No extra dependencies. The one thing that trips people up is expecting the gallery to behave like a plain file browser - it isn't one. The image_paths string is managed by the node's UI; don't hand-edit it. If your images aren't showing up, check you're in the right input subfolder and that the folder picker (not just the upload button) is what you used. And remember: everything gets resized to one batch size, so if you mix portrait and landscape references, Center Crop (Fill) will decide what survives - that's a feature, but it's a decision you should make on purpose.
Inputs (9)
| Name | Type | Default | Description |
|---|---|---|---|
| image_paths | STRING | Selected image file list managed by the node UI. | |
| mode | COMBO | Keep Input Ratio | Choose how each image is resized before the batch output is built. |
| ratio_preset | COMBO | 16:9 | Target aspect ratio used in Preset Ratio mode. |
| megapixels | FLOAT | 1.000.01–10 | Target total image size in megapixels for automatic sizing modes. |
| width | INT | 102464–8192 | Manual or fallback output width in pixels. |
| height | INT | 102464–8192 | Manual or fallback output height in pixels. |
| divisible_by | COMBO | 32 | Round the final batch size to a model-safe multiple. |
| interpolation | COMBO | lanczos | Resize filter. Lanczos is the default quality choice. |
| resize_method | COMBO | Center Crop (Fill) | Choose crop-fill or fit/letterbox behavior for loaded images. |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| multi_output | IMAGE | Loaded images resized into one same-size IMAGE batch. |
| width | INT | Final batch width in pixels. |
| height | INT | Final batch height in pixels. |