Ino Validate Media Files

Every dataset has that one file. The half-downloaded JPEG, the video that's actually a text file renamed to .mp4, the PNG that opens in nothing. Usually you find out halfway through a training run or a batch process, and the error message points at a file you have to hunt down manually. Ino Validate Media Files sweeps a folder for you, sorts the good media from the bad, and hands back a report - the cleanup tool for anyone scraping datasets or ingesting folders of files.

It's part of the ComfyUI Ino Nodes pack (nobandegani/ComfyUI-InoNodes), in the file-helper family.

What it does

Point it at a folder (via the pack's standard parent_folder + folder inputs) and it scans the contents, categorizing files as valid images, valid videos, or problems - separating unsupported formats from files that fail validation. It works off the pack's shared file helper, and its output is a small report card rather than a single value.

Inputs that matter

• enabled - on by default. Off means it reports failure without scanning.
• parent_folder - input, output, or temp as the root.
• folder - the subfolder to scan.
• include_images and include_videos - toggles that let you check only one media type, on by default. If you only care about videos, switch images off and it won't spend time on them.

Outputs

It's output-heavy, and that's the point - each problem category gets its own path:

• success (boolean) and message (string) - did the scan complete.
• rel_path / abs_path - where you pointed it.
• skipped_images_path, skipped_videos_path - where failed images/videos ended up.
• skipped_images_unsupported_path, skipped_videos_unsupported_path - unsupported-format files.
• unsupported_files_path - anything else that isn't recognized media.
• logs - a text log of what happened.
• output_path - the base output location.

Wire the path outputs into a zip node or a remove-file node and you can act on the results automatically, not just read them. The practical pattern: run validation on a freshly scraped folder before a training or batch step, and let the report tell you exactly what's broken.

Install

# ComfyUI Manager: search "ComfyUI Ino Nodes", install, restart
# or manually:
cd ComfyUI/custom_nodes
git clone https://github.com/nobandegani/ComfyUI-InoNodes.git
cd ComfyUI-InoNodes
pip install -r requirements.txt

Restart ComfyUI after. No model downloads - the heavy lifting lives in the pack's shared dependency inopyutils, installed by requirements.txt. The pack targets ComfyUI's V3 schema, so keep ComfyUI current (v0.18.1+); and the README's cd comfyui_ino_nodes is a typo for the actual ComfyUI-InoNodes folder.

The one thing to expect: validation takes as long as reading every file takes, so a big folder isn't instant - that's the cost of actually checking files rather than trusting extensions. Worth it the first time it catches a corrupt file that would have killed your next run.