📥 Load Pinterest

Pinterest is where half the internet keeps its mood boards, and this node pipes it straight into ComfyUI. Give it a Pinterest link - a single pin, a board, a board section, a whole profile - or just a search keyword, and it downloads images from there and hands them to your workflow. It's built for reference gathering: pulling style references for IPAdapter, building a quick dataset, grabbing a batch of "this kind of thing" to run through img2img.

The keyword mode is the fun part. Type vintage film photography and it searches Pinterest and pulls matching images, no link required. It's a reference-scraper with generation attached.

How it works

Under the hood it uses gallery-dl (the same scraper the pack's other web-loaders lean on) to fetch images and save them into your input folder, organized into their own subfolders so a big pull doesn't clutter everything. It downloads a range of images, then loads a selection of them back as an IMAGE list. The README notes a couple of niceties: if your URL starts with a slash it auto-prepends https://www.pinterest.com, so /user/board expands to the full link, and if you pass a keyword instead of a URL it searches.

There's a two-step count going on, which trips people up if they don't clock it: range controls how many images get downloaded to disk, and number controls how many of those get loaded into the workflow. So you might scrape 50 to your input folder but only pull 5 into the graph.

The inputs that matter

• url - a Pinterest link (pin, board, section, or profile) or a search keyword. Both work; a leading slash gets the domain auto-added.
• range - the location/count of images downloaded to the input folder (default 1-10). This is your scrape size.
• number - how many images to actually load and return downstream.
• random - if true, the loaded images are picked randomly from what was downloaded; if false, they're chosen by position (driven by the seed).
• seed - makes the non-random selection (and reproducibility) deterministic.

The single output is IMAGE, a list - wire it into whatever's consuming your references.

Common issues

It downloads nothing. Pinterest scraping is fragile by nature. Private boards, an unreachable profile, rate-limiting, or gallery-dl needing an update all cause empty pulls. Try a public board or a plain keyword search first to confirm the pipe works, then narrow down.

Fails on Windows or macOS. The download side needs aria2c, which isn't installed by default on those systems. Install it manually - the README calls this out for every auto-download node in the pack.

I asked for 10 but got fewer. range and number are separate dials. If range downloaded fewer images than number wants to load (or the search just didn't return enough), you get what's available. Widen range.

The images are lower quality than the pins looked. Pinterest serves multiple sizes; the scraper grabs what it can resolve. The README mentions it tries to find the highest-quality version, but you're at the mercy of what the source hosts.

A fair word: scraping other people's images is for reference and personal experimentation - mind the usage terms of anything you pull, especially before you build a training set out of it.

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. Install aria2c on Windows/macOS for the download to work. The node is under 📂 SDVN.