🏞️ Load Image Ultimate

This is the kitchen-sink image loader - one node that swallows five different ways of getting images into a workflow and lets you switch between them with a dropdown. Input folder, a custom folder path, a direct URL, Pinterest, Instagram. The idea is that instead of swapping loader nodes when your source changes, you keep this one wired in and just flip mode. The fields it shows adjust to whatever mode you picked, so you're not staring at ten irrelevant inputs.

It's convenient and it's a little overwhelming, both true. If you know you're only ever loading from one source, the dedicated single-purpose loaders (Load Image Folder, Load Pinterest) are cleaner. But if you build workflows you reshare, or you genuinely bounce between local files and web sources, having one loader that does it all is the whole point.

How it works

Pick a mode and the node behaves like the matching dedicated loader - the README says the usage is "similar to the corresponding image loading nodes above," because it literally is the same machinery collapsed into one node. Local modes read from disk; the URL mode downloads a linked image; Pinterest and Instagram modes use gallery-dl to scrape and download images to your input folder, organized into subfolders, then load a selection back. Images come out as a list so downstream nodes iterate over them.

The inputs that matter

You won't touch most of these at once - they light up per mode. The ones that matter by mode:

• mode - the source: Input folder, Custom folder, Url, Pintrest, or Insta. This is the master switch.
• image - the file picker for Input folder mode.
• folder_path - the directory for Custom folder mode.
• url / insta_url / pin_url - the link for Url, Insta, and Pinterest modes respectively.
• range - for the scraping modes, how many images to download to the input folder (default 1-10).
• number / number_img - how many images to actually load and return.
• random + seed - random selection vs. deterministic-by-position, with a seed for reproducibility.
• index - position selection where relevant.

Outputs are IMAGE (a list) and MASK - the mask being useful when a loaded image carries an alpha channel.

Common issues

Pinterest/Instagram modes fail or return nothing. Those lean on gallery-dl to scrape, and web scraping is inherently fragile - sites change layouts, rate-limit, or require the target to be public. If a board is private or the site pushed back, you get nothing. Also, on Windows/macOS you need aria2c installed for the download side.

The node is a wall of inputs and it's confusing. That's the trade for one-node-does-everything. Set mode first; ignore the fields that don't belong to your mode. If it still feels heavy, use the single-purpose loaders instead.

Custom folder returns empty on a remote host. folder_path is server-side - the path has to exist on the machine running ComfyUI, not your local desktop.

"Pintrest" is spelled wrong in the dropdown. Yep, it's a typo in the node's mode list. Pick it anyway; it works.

Installing SDVN_Comfy_node

ComfyUI Manager: search "SDVN_Comfy_node" → install → restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/StableDiffusionVN/SDVN_Comfy_node

Then pip install -r custom_nodes/SDVN_Comfy_node/requirements.txt from the ComfyUI root and restart. For the Pinterest/Instagram/URL modes, install aria2c on Windows or macOS. The node is under 📂 SDVN.