Load Bytes (Yogurt Nodes)

Some files aren't images, masks, or audio. They're just bytes - a JSON file, a binary blob, a serialized object - and if you want a ComfyUI workflow to read one, you need a node that will shut up and hand you the raw content. YogurtLoadBytes does that. One path in, one BYTES output, no interpretation, no surprises.

It's part of ComfyUI-YogurtNodes, yogurt7771's 150+ node all-in-one pack. The pack is obscure and MIT-licensed with an auto-generated README, but its IO section is built around exactly this kind of "boring plumbing" - and boring plumbing is what makes complex automation graphs possible.

How it works

A single required input, file_path - a STRING with the path to the file. The output is bytes_data, typed BYTES, containing the file's raw contents read straight off disk.

On its own, raw bytes aren't much use. The value appears when you pair it with the pack's other IO pieces:

• Serialize Any / Deserialize Any - the pack can pickle Python objects to bytes and back. YogurtLoadBytes feeds a deserializer with a previously saved object, letting you persist data between runs or across workflows.
• Save Bytes Bridge - writes bytes back out to your ComfyUI output directory, closing the read-modify-write loop.

So the realistic use is a data round-trip: serialize some state, save it, load it again later, deserialize. Or simply reading arbitrary files that a workflow needs to inspect - JSON configs, downloaded blobs, anything a downstream API or script node wants in hand.

Where you'll use it

• Reading a JSON state file and piping it into YogurtJsonParse for processing.
• Restoring a pickled object saved by YogurtSerializeAny on a previous run.
• Feeding raw file content into any node that expects BYTES, without the pack guessing at a format.

The main thing to keep straight: this is a raw read. It does not detect format, does not decode, does not care what the file is. If you hand it a path that doesn't exist, the graph errors at queue time, so it's worth checking paths upstream before you run a long batch. And because it reads the whole file into memory, giant files will eat RAM - fine for configs and objects, less fine for multi-gigabyte dumps.

Install

ComfyUI Manager, search ComfyUI-YogurtNodes, install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/yogurt7771/ComfyUI-YogurtNodes
cd ComfyUI-YogurtNodes
pip install -r requirements.txt

Restart, look under "Yogurt Nodes". Light dependencies, no model downloads - nothing about this node is exotic, and that's the appeal.