Social Media Downloader (gallery-dl)

The Social Media Downloader (GalleryDLDownloader, part of the Download Tools pack) is the node you reach for when you want every image from an Instagram profile, a Reddit sub, a Pixiv artist, a Twitter account, a DeviantArt gallery - 100+ sites' worth of archives - sitting in a folder on your machine. The scraper node in the same pack pokes at whatever a page happens to render; this one calls in the specialist that does it properly.

It's a thin, honest wrapper around gallery-dl, the battle-tested Python downloader by Mike Fährmann that people have been using for years from the command line. The node builds a gallery-dl command from your settings and shells out to it. No reinvented logic to get subtly wrong, no missing corner cases - the entire 100-site feature set of gallery-dl is yours, just behind a ComfyUI node. For bulk archiving this is the one you'd actually reach for.

How it works

Paste URLs into url_list (one per line, # lines are ignored), pick where files go, hit run. The node invokes gallery-dl with your options: a per-site config file, optional browser cookies, an SQLite download archive that tracks what's already saved, resolution filtering, and a timeout. Files land organized into output_dir, sorted by site and user, and optionally into images/ / videos/ subfolders.

The inputs that matter

• url_list and output_dir - required, self-explanatory.
• config_path - the key one. A single gallery-dl.conf holds credentials for every site (Instagram, Reddit, 500px, DeviantArt, Pinterest, Flickr, Bluesky) and gallery-dl auto-selects the right ones from the URL. Point it at ./configs/gallery-dl.conf and the path is remembered between sessions.
• filter_by_resolution (on by default) with min_image_width / min_image_height - skips thumbnails and icons under 768px. Videos are never filtered. Note the honest caveat in the tooltip: sites without dimension metadata (Bunkr, Cyberdrop) download everything regardless.
• instagram_include - posts, stories, highlights, reels, or comma-combined. Instagram-only, and the reason the node's README leans so heavily on Instagram.
• download_timeout - default 1800s (30 min). Huge galleries can need hours. When it times out you just run again: the archive means it resumes where it stopped.
• use_browser_cookies / browser_name - leave off, per the README; the config file is more reliable than cookie-jacking.

The extra_options field is a backdoor to the full gallery-dl CLI - --limit 50, --range 1-100, --ugoira-conv for Pixiv animations, -o directory=[...] for custom folder layouts. If you know gallery-dl, you know this field.

The outputs

output_dir (STRING) - where things went, wire it anywhere you need the path; summary (STRING) - a human-readable report; download_count (INT); and success (BOOLEAN) for routing in the graph. This node doesn't feed the sampler - it's a data-acquisition step you run to build a dataset or reference folder.

Installing it

ComfyUI Manager, search "Download Tools." Or:

cd ComfyUI/custom_nodes
git clone https://github.com/EricRollei/Download_Tools
cd Download_Tools && pip install -r requirements.txt

gallery-dl installs as a Python dependency - no separate binary, no browser download. FFmpeg is only needed for niche extras like ugoira conversion.

Where it bites

• 401 Unauthorized on Instagram - expired session cookies (they die after roughly a year). Export fresh ones via the "Get cookies.txt LOCALLY" browser extension into the config file.
• Chrome cookies unreachable - Chrome locks its cookie DB while running; close it, run ComfyUI as admin, or just use Firefox.
• Too many small images - that's the resolution filter being off, or a site that reports no dimensions.
• Timeout - bump download_timeout (up to 10 hours) and re-run; it resumes.

One word of caution, since this node's whole job is downloading from sites that often frown on it: it's a non-commercial (CC BY-NC 4.0) pack with a separate commercial license, and you're responsible for what you archive and how you use it. For your own content and clearly okay-to-keep material, it's superb.