Install & quickstart

By the end of this page you’ll have generated an image using a StyleRef style, without creating an account.

Install

Using ComfyUI Manager (recommended):

1. Open Manager → Custom Nodes Manager.
2. Search for StyleRef.
3. Click Install, then restart ComfyUI.

Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/StyleRef/styleref-comfyui

Restart ComfyUI.

There is nothing else to install — the nodes use only the Python standard library. On a healthy start, the server console prints a line like [StyleRef] v0.1.0 — 5 nodes registered.

Generate your first styled image

1. Add the Load node

Double-click the canvas, search for StyleRef Load, and add it.

In the style_ref field, paste a style slug. To try one immediately:

9a2adtz6-cd8d77ee2f51

Any style from the gallery  works — open one and copy the slug from its URL.

Don’t know the slug? Click Search styles… on the node — a picker opens with live search, category filters, and thumbnails; when you’re signed in, your own styles are listed on top. Picking one fills style_ref for you.

No picker button (headless frontend, older install)? Type a query like warm editorial into the search field and queue the Load node — matching styles, with their slugs, appear right in the node body. Queuing just this node performs no generation.

2. Add the Apply node

Add StyleRef Apply and connect style from Load into it.

Then:

• subject — what to depict, e.g. a lighthouse on a rocky shore at dawn
• target — the format to compile, same list as the web app. For the sampler in your graph: flux for FLUX models, diffusion for SDXL/SD1.5. (ai_tools, style_md, midjourney, and json are for copying the output out of ComfyUI.)

The subject describes what; the style describes how it looks.

3. Wire it into your sampler

Connect Apply’s outputs to your two CLIP Text Encode nodes:

• positive → your positive prompt encoder
• negative → your negative prompt encoder

The negative output is the style’s own guardrails (plus anything you type in extra_negative).

4. Queue it

That’s a complete graph. Queue the prompt.

After the run, Apply’s node body shows the composed positive prompt and a rough token estimate — what StyleRef actually sent to the encoder.

To change the subject, edit the subject field and re-queue — the style stays fixed. That’s the point: one style, many subjects, a coherent set.

Start from a template instead

If you’d rather not wire it by hand, drag 01-quickstart.json onto the canvas — it’s the graph above, complete with a preview of the compiled prompts. Set your checkpoint and queue.

See workflow templates for all four.

Next steps

• Nodes reference — every input and output
• Login & credits — use your own styles, including ones you extracted on styleref.io

Troubleshooting

The nodes don’t appear after installing. Restart ComfyUI fully, then check the server console for the [StyleRef] v… — 5 nodes registered startup line. If it’s absent, the pack didn’t import — the console will say why near the import errors.

“Could not reach styleref.io”. The machine running ComfyUI needs internet access — which is not always the machine with the browser. On a remote or containerized install, check the container’s outbound network.

Rate limited. Anonymous requests are rate-limited per IP. Wait a moment and re-queue; the node caches loaded styles, so re-running with the same style doesn’t re-fetch it.