Load Video by Path

This is the smallest node in the pack, and it exists for one reason: Convert Video (FFmpeg) re-encodes much faster when it can touch the file directly instead of eating a decode-to-frames round trip. Load Video by Path is the feeder that makes that fast path possible. You probably won't use it on its own - it's a picker, not a processor - but the pair of them is the point of this whole pack.

How it works

It scans your ComfyUI/input folder for video files, filters to the ones FFmpeg-class loaders recognize, and gives you a dropdown of what it finds. Pick one and it returns a VIDEO output - but not a decoded one. Behind the scenes it wraps the file path in a little marker object that tells the converter "this is already a path, don't touch it," which is how the convert node knows to skip the frame-by-frame detour and pass the path straight to FFmpeg.

That's the whole mechanism, and it's the whole value: a re-encode that would mean dumping hundreds of PNGs to disk and re-encoding them from scratch instead becomes a direct file operation. Order of magnitude faster on anything long.

What you actually set

One input, one output - this is the leanest node you'll meet.

• video_file - a dropdown of video files in your input directory. There's also an upload widget, so you can drag a clip in from the browser and it lands in input/ ready to pick.
• VIDEO - the output, and here's the catch: it's only compatible with the Convert Video (FFmpeg) node from this same pack. Don't try to wire it into the standard video-loading path or frame-based nodes; it won't behave like a decoded video.

When to reach for it

If your workflow is "generated clip → want it smaller / WebM / without audio," wire Load Video by Path into Convert Video and let the fast path do the work. If you actually need the frames - analysis, per-frame upscaling, feeding a video model - this is the wrong tool; use ComfyUI's normal loader, which gives you real frames. This one is a courier, not a decoder.

Installation is shared with the rest of the pack: ComfyUI Manager, search "Save Images to Video", or clone https://github.com/San4itos/ComfyUI-Save-Images-as-Video into custom_nodes, run pip install -r requirements.txt (only Pillow - no models, no heavy deps), and restart. And remember the pack-wide requirement: FFmpeg itself has to be findable, via ffmpeg_config.ini, the ffmpeg_bin/ folder, or your system PATH. It's a two-node friendship - picker and converter - and neither works without the binary standing behind them.