Load Character

The point of saving a character is being able to load it back. CP_LoadCharacter is the reload: it drops a dropdown of every .char file the pack can find and hands you the CHARACTER object, ready to feed a render, a repair, or a caption node. If you've been encoding and re-encoding the same hero every session, this is the node that makes that habit obsolete.

How it works

On graph execution it scans the character output directory (ComfyUI/output/characters/) for .char packages and lists them in the character_file dropdown. Pick one, and the node reads the bundle back: reference image, palette, embedding, and all the metadata (captions, prompt templates) that CP_CharacterCaption and CP_CharacterEncode stashed in there.

One input - character_file, a dropdown - and one output - character.

Installing

Part of Spit8/_ComfyUI_CharacterPose:

cd ComfyUI/custom_nodes
git clone https://github.com/Spit8/_ComfyUI_CharacterPose
pip install -r _ComfyUI_CharacterPose/requirements.txt

Restart ComfyUI; it's under CharacterPose/Character.

Gotchas

• The dropdown only refreshes when the graph loads. Save a new character mid-session and it may not appear until you reload the workflow. Annoying, standard ComfyUI dropdown behavior, not a bug.
• If nothing's saved yet, the dropdown holds a (none) entry and the node raises a clear error: "No .char files found. Encode and Save a character first." That's the pack telling you the exact next step.
• A .char is a folder (meta.json + tensors.safetensors inside). It's portable - copy the whole folder to another machine's output/characters/ and it loads there too.
• Loaded characters carry their cached captions, which is exactly what CP_CharacterCaption uses when use_cached_caption is on. So a loaded character skips the captioner entirely on subsequent runs.

Encode, save, close ComfyUI, come back a week later, load, keep going. That's the workflow this node exists to enable - and it does it with a single dropdown.