Prompt Styler Camera

Most nodes in wolfden's SDXL Prompt Styler pack style your prompt by art movement or scene. This one styles it by camera body - Canon EOS R5, Nikon, Fuji, Sony, the whole spread, ~127 entries. You type your subject, pick a camera off the dropdown, and it wraps your text in that camera's descriptor. It's a shortcut to photographic realism without having to know which bodies and lenses the model responds to.

And here's the interesting part, because it's not just convenience. Naming a real, physical camera is one of the highest-leverage things you can put in a prompt on any encoder trained on real photo captions - it drops the model out of its glossy "beauty default" faster than piling on words like realistic, candid, or amateur ever will. So of everything in this pack, the camera node is one of the few whose trick genuinely carries over to newer models, not just SDXL.

How it works

Each camera is a JSON entry with a prompt field holding a {prompt} placeholder and a matching negative_prompt. Pick a body, and the node substitutes your text_positive into {prompt}, producing something like photo taken with a Canon EOS R5 of a portrait of an old fisherman .... No model, no API, no key - it's plain string substitution. The value is entirely in that camera name being a concrete referent the encoder recognizes.

The negative prompt behaves like the rest of the pack: the preset ships one, and if you also type into text_negative the two are merged.

The inputs that matter

• text_positive - your subject. Keep it clean; the camera descriptor does the photographic lifting.
• style - the camera body, ~127 choices. Start typing the brand to filter.
• text_negative - optional, merged with the preset's negative.

log_prompt (No/Yes) prints the assembled prompt to the console - worth flipping on once so you see exactly what descriptor a given body adds. Outputs are positive_prompt_text_g and negative_prompt_text_g, both strings; feed positive into the CLIP Text Encode driving your positive conditioning and negative into the negative one. On SDXL the same string can populate both G and L of a CLIPTextEncodeSDXL.

Installing it

ComfyUI Manager: search SDXL Prompt Styler (wolfden), install, restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/wolfden/ComfyUi_PromptStylers

Restart ComfyUI and look under the new Style Prompts submenu. No downloads, no dependencies - string handling only.

Where people get tripped up

The camera name is the leverage, but the node bolts on the whole preset descriptor, which is sometimes more marketing-speak than you want. If the shot feels over-styled, turn on log_prompt, read what it appended, and consider whether you'd rather just type photo taken with a Canon EOS R5 yourself and skip the extra adjectives. The presets are JSON in the pack folder - easy to open and trim.

Also keep the negative in mind. On SDXL and its lineage the preset negative is fine, but on a distilled model running at CFG 1 (any Turbo/Lightning checkpoint) the negative does nothing at all, since CFG 1 skips the negative pass entirely. That doesn't hurt the camera trick - the positive camera name still lands - it just means the negative half of the node is dead weight there.

If the node never shows up, you skipped a full restart or you're hunting in the root menu instead of Style Prompts. Of the whole pack, this is the one I'd keep in reach even on modern models - the camera name is doing real work.