*️⃣ Model info editor

If you've ever downloaded a LoRA six months later and had no idea what trigger word it needed or where you even got it, this is the node that prevents that from happening to someone else - or to future you. It writes metadata straight into a checkpoint or LoRA file: trigger word, copyright, source URL, notes, even a cover thumbnail, so the file carries its own documentation instead of relying on a CivitAI page that might not exist anymore.

How it works

.safetensors files aren't just tensor data - the format starts with a JSON header before the weight bytes, and that header can hold arbitrary metadata alongside the tensor shapes. This node writes your trigger word, copyright line, and source URL into that header when Json_Embed is on, so the information travels with the file itself: rename it, move it between machines, share it with someone else, and the metadata comes along. Separately (or additionally), it can drop a plain .txt sidecar next to the model with your longer notes, for anything you don't want baked into the header itself - or for a quick reference you can open without any tooling at all.

This matters more than it sounds like it should. Community habit is to track trigger words and source links through the model's CivitAI page or the file's original folder structure, and both of those degrade over time - pages get taken down, folders get reorganized, and a model that's actually excellent becomes a file nobody can use correctly because nobody remembers the magic word it needs.

The inputs and outputs that matter

• model_path - the file to edit.
• Json_Embed (default true) - write the fields below into the file's own safetensors header.
• Trigger_word - the word or phrase the model needs in the prompt to activate its trained concept, if it has one.
• Copyright - a credit or licensing line.
• Url - where the model came from, or where to find more info about it.
• Txt_save (default true) - additionally write a .txt sidecar file next to the model.
• Txt_note - freeform notes for that sidecar.
• image_cover (optional, IMAGE) - a preview thumbnail to attach.

The output is model_path, echoing the path back out - useful for chaining this node into something that acts on the same file afterward (for example, the pack's Model Export node).

Installing it

ComfyUI Manager: search SDVN_Comfy_node, install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/StableDiffusionVN/SDVN_Comfy_node

Then pip install -r custom_nodes/SDVN_Comfy_node/requirements.txt from your ComfyUI root, and restart.

Where people get burned

This edits the file in place. model_path is both the input and, implicitly, the target - there's no separate output path field. Keep a backup of anything valuable before running this against it, the same way you would before any tool that rewrites a multi-gigabyte file.

Embedded metadata isn't universally read by every tool. Writing your trigger word into the safetensors header is genuinely useful for anyone loading the file through this pack's own info-reading nodes (or anything else that bothers to check), but not every loader or platform surfaces custom header fields to a user - the .txt sidecar is the belt-and-suspenders option precisely because it's readable by literally anything with no parsing required.

Blank fields don't clear existing metadata by themselves - verify what actually landed. If you're re-running this on a file that already has header metadata from a previous editing pass, check the result rather than assuming an empty field means "no change."