Prompt Styles v2

"Prompt Styles v2," as this node is labelled, is Searge SDXL's take on saved prompt styles - the idea that instead of retyping "cinematic, dramatic lighting, 35mm" onto every prompt, you pick a named style and it gets applied for you. It's one of the input stages that rides on the Magic Box data stream, and it's a feature that took a while to actually land in this pack.

Worth knowing the history, because it's right there in the README. When the v4.0 rewrite shipped, the author listed prompt styling as one of the things that got cut to release earlier - "the ability to load styles from a template file and apply them to prompts." The "v2" in the name, and the fact that this is a proper Searge/UI/Inputs node rather than the old [unfinished work in progress] style preprocessor, is the sign that the feature eventually got built out.

How it works

Like everything else in the v4.x workflow, this node sits on the single SRG_DATA_STREAM that carries the pipeline's state from stage to stage. It reads the stream in, applies the selected style to the prompt data travelling on it, and passes the stream along. The style selection itself is a UI concern on the node - you choose a style, and the pipeline folds it into the prompts before they reach the sampler, so you don't have to paste style boilerplate into your text inputs by hand.

That's the whole value proposition: reusable, named prompt modifiers, managed in one place instead of copy-pasted. It's the same instinct behind style pickers in other UIs, wired into Searge's data-stream architecture.

The inputs and outputs that matter

The node's data contract is deliberately minimal. The one connection you make is the stream:

• data (optional, SRG_DATA_STREAM) - the incoming Magic Box stream from the previous stage.
• output data (SRG_DATA_STREAM) - the stream passed onward, now carrying the applied style.

Everything else about this node - which styles exist, which one is active - lives in its on-node UI rather than as wired inputs, which is why the connection surface is just the stream in and the stream out. You place it in the pipeline, pick a style, and let the data stream carry the result.

How to install it

Comes with the Searge SDXL pack. ComfyUI Manager: search SeargeSDXL, install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/SeargeDP/SeargeSDXL.git

then restart. As with the rest of the pack, manual installers need python -m pip install opencv-python run once in ComfyUI's Python environment or it won't load.

Common issues

Because this node only speaks SRG_DATA_STREAM, it's only useful inside the Searge Magic Box - it's not a general-purpose styler you can bolt onto some other workflow. Use it in place, in the stock Searge graph, where the stream is already flowing. And keep in mind styles are additive: a heavy style plus an already-detailed prompt can fight each other, so if output drifts away from what you asked for, try a lighter style or none. The usual pack-wide caveat applies too - if the node errors after a ComfyUI update, update Searge first (the v4.3.x releases were compatibility-only), and always pair the latest workflow JSON with the latest nodes.