⏬️ Image Scraper

Image Scraper pulls images off a website straight into ComfyUI, using gallery-dl under the hood. Point it at a gallery, a board, a profile, or a search page, and it downloads the images to a folder and hands them to your graph as a list. If you're building a dataset - reference images for a LoRA, a batch of style examples, a mood board to run through img2img - this is the node that saves you from right-click-saving a hundred images by hand.

gallery-dl is the important detail, because it's what gives this reach. It's a mature command-line scraper that supports a huge list of sites (the README links its supported-sites list), so this isn't a Pinterest-only or Danbooru-only tool - it handles boards, tags, user galleries, and search results across many platforms, resolving them down to the actual full-resolution image files. The node wraps that and feeds the results back as an IMAGE list.

The inputs are few and practical. url is the page you're scraping. save_folder picks whether the files land in your ComfyUI input or output directory, and custom_save_folder lets you drop them in a named subfolder so a scrape doesn't get lost in the pile. The one input worth understanding is cookies: set to chrome or firefox, it borrows your browser's cookies so gallery-dl can see content that requires being logged in - age-gated pages, follower-only galleries, anything behind a session. The single image output is the list of everything it pulled, and this is an output node, so it'll run and save on its own even if you don't wire the images onward.

Where it shines is dataset-building and bulk reference collection, and the way you'd actually use it is scrape-to-input, then feed the list through a filter or a resize and into whatever you're training or generating from. Because it's built on gallery-dl rather than a bespoke scraper, it tends to keep working on sites where a hand-rolled downloader would break, which is the whole reason to prefer a wrapper around a maintained tool over reinventing one.

Install comes with the pack. ComfyUI Manager, search SDVN_Comfy_node; or 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 your ComfyUI root and restart - gallery-dl comes in through that requirements install. Downloads also lean on aria2c for speed, so Windows and macOS users should install aria2c by hand; on Linux it's typically already there.

Troubleshooting is mostly about access and the tool underneath. If a scrape returns nothing, the two usual causes are a site gallery-dl doesn't support (check its supported-sites list) or content that needs a login you didn't provide - that's the cookies input's whole job, so set it to the browser where you're actually signed in to that site. If it worked yesterday and broke today, that's often gallery-dl needing an update as a site changed its layout; updating the package under the pack usually fixes it. And a reminder that isn't about the node so much as the material: scraping pulls down whatever's on the page, so mind the licensing of what you collect, especially if a LoRA trained on it is going anywhere public. The node makes the collecting frictionless; what you do with the pile is on you.