Load Media (img, vid or dir)

Load Media is the pack's flagship loader, and the reason it exists is written in its own source: it "steals from all the major image loader nodes." Point it at a file, a folder, a video, a zip/tar/7z archive, a URL, or even a wildcard pattern like *.png, and it loads whatever's there as a clean image sequence - with metadata, timing, and sorting options most loaders don't bother with. If you do frame-based work (video, animation, batch processing), this replaces half your loader collection.

What it takes in

The path input is the whole show - an image, a directory, a video file, an archive, a URL, or a glob. The rest are sequence controls:

• sort - None, alphabetical, date_created, date_modified, visual_path, or random. The interesting one is visual_path: it orders frames by visual similarity using a traveling-salesman-style solver (borrowed from the community's sort_visual_path work), which is a game-changer for sequences whose filenames lie about their order.
• start_index, skip_n, image_load_cap - slice the sequence any way you want.
• start_index_use_seed + seed - use the seed as the starting frame for random/sequential runs (great for reproducible "which frame did I start on" workflows).
• loop_first_frame - appends the first frame to the end for seamless loops.
• error_after_last_frame - raise instead of wrapping around with modulo when you overshoot.
• resize_images_to_first - force every image to the first frame's dimensions (batch-friendly, and it's on by default).

Outputs are the full spread: image, mask, WIDTH, HEIGHT, COUNT, FILE_NAME, FILE_PATH, PARENT_DIRECTORY, FPS, AUDIO, PROMPT, and METADATA_RAW. The FPS + AUDIO pair is what makes video work flow - you can hand both to a video-combine node and keep timing and sound without a second loader.

Video, archives, URLs

For video it detects ffmpeg and extracts frames (duration and rate come back as FPS and the audio track as AUDIO). Archives (zip/tar/7z) get unpacked into a temp dir and treated as an image sequence - handy for batch-downloaded datasets. URLs are fetched over HTTP. The one honest caveat: this is a big node with a big requirements.txt behind it (moviepy, py7zr, librosa, opencv…), so if you only need a plain image loader it's overkill. It earns its place when you're juggling mixed media - the "one loader for everything" promise is real, and it's the node people keep in their video workflows for exactly that reason.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/drmbt/comfyui-dreambait-nodes

then restart ComfyUI (or ComfyUI Manager, search "comfyui-dreambait-nodes"). First run with a video needs ffmpeg available on PATH; if frame extraction silently fails, that's the first thing to check. It's a heavy install for a loader, but for animation and video work it's one of the better "just load the media" nodes around.