Image to Seed

This is a niche little utility with one clear job: take an image in, get a deterministic seed number out. The README puts it plainly - "convert an image to a reproducible seed." Feed it the same image and you always get the same number back. Feed it a different image and you get a different one.

Why would you ever want this?

It sounds pointless until you hit the use case, then it's exactly right. The core idea is deriving a seed from image content instead of picking one at random or typing one in. A few places it earns its keep:

• Content-linked reproducibility. You want a generation whose seed is tied to a specific input image, so that "same input → same result" holds without you tracking seeds in a spreadsheet. Great for pipelines where an image drives a downstream generation and you want the pairing to be repeatable.
• Seeding variation off a batch. Walking through a folder of images and want each one to deterministically produce its own seed for a paired output? This gives you that mapping for free.
• Turning "this image" into a stable handle. Anywhere you need a number that uniquely-ish represents an image and stays constant.

It's the kind of node you don't need until you're building something a bit clever, and then it's the missing piece.

How it works

Under the hood it hashes the image's pixel data down to an integer in the seed range. Same pixels, same hash, same seed - that's the "reproducible" guarantee. The input is an image; the output is a seed number you wire into a KSampler's seed input or any node that takes a number.

The consequence to internalize: it's a hash, so it's sensitive. Re-encode the image, resave it as a different format, resize it by a pixel, run it through a filter - the pixels change, and the seed changes completely. There's no "close images get close seeds"; a hash doesn't work that way. So if you're relying on this for reproducibility, freeze the exact image bytes you feed it. Two visually identical PNGs that went through different save paths can hand you two different seeds.

Installing it

Comes bundled in WAS Node Suite. Install the pack via ComfyUI Manager - search was-node-suite-comfyui, install, restart - or clone manually:

cd ComfyUI/custom_nodes
git clone https://github.com/WASasquatch/was-node-suite-comfyui

then install requirements.txt and restart. No models involved; it's a pure computation.

Troubleshooting

The pack itself is the thing that tends to break, not this node. WAS Node Suite has been unmaintained since December 2023, and the recurring failure is the whole suite reporting "Import Failed" after a ComfyUI update - a dependency-version clash (opencv is the usual suspect, since WAS pins an older release). The fix is reinstalling requirements.txt into the correct venv (activate it first, or run the pack's install.bat). Once the suite imports, Image to Seed is deterministic and dependable - that's its entire point.