DP Prompt Styler

Type your subject once, then dress it with dropdowns. This node takes your core idea and layers on the cinematography vocabulary that's tedious to remember - depth of field, camera angle, color theme, lighting, time of day, mood, film-camera model - and glues it all into one finished prompt string. It's the difference between writing "portrait of a woman" and getting "portrait of a woman, shallow depth of field, low angle shot, warm autumn earth tones, golden hour, captured on a Canon EOS R5" without touching the keyboard nine times.

It's the flagship text node in DesertPixelAi's Desert Pixel pack, and it's the one to reach for when you want consistent, photographic-sounding prompts across a series.

How it works

You put your idea in Main_Prompt, and then set up to nine style categories, each a dropdown: DepthStyle, cameraAngles, colorTheme, FaceMood, timeOfDay, atmosphere, lighting, filter_effect, and CameraType. Every category has dozens of options and two special entries at the top - none (skip this category) and randomize (roll this one category each run). The node assembles your main prompt plus the chosen phrases into a single Modified Prompt output.

The mode switch sits over the top:

• Styler_ON - apply the categories as you've set them.
• ByPass_All - pass your prompt straight through untouched, no styling. A clean on/off without unplugging anything.
• Randomize_All - roll every category at once. Slot-machine cinematography.

The optional text slots - pre_text, pre_text_B, extra_text, extra_text_B - let you inject your own fragments before and after, so you're not limited to the presets.

The inputs and outputs that matter

Main_Prompt (your subject), mode (on / bypass / randomize), and whichever of the nine category dropdowns you care about - leave the rest on none. The single output, Modified Prompt (STRING), goes into your positive CLIP Text Encode.

The habit that pays off: set the categories that define your series (say, lighting and colorTheme) to fixed values, and set one or two others to randomize for controlled variety. You get a coherent look with just enough movement.

How to install it

ComfyUI Manager: search Desert Pixel, install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/DesertPixelAi/ComfyUI-Desert-Pixel-Nodes

Restart ComfyUI. No downloads. Under the DP category.

Common issues & troubleshooting

The prompt got long and the subject drifted. Stacking nine categories piles on tokens, and past a point the model starts ignoring the tail. Use fewer categories, or lean on pre_text/extra_text to keep your subject words prominent rather than buried behind style phrases.

Styling seems to do nothing. Check mode - if it's on ByPass_All, every category is ignored by design. Switch to Styler_ON.

It changes every run and you can't compare. A category set to randomize, or mode on Randomize_All, re-rolls each queue. Pin the categories to fixed values when you're tuning other settings.

A style term fights your checkpoint. Photographic terms like "captured on a Leica" push hard toward realism; on a stylized/anime model they can muddy things. Drop CameraType and filter_effect on non-photo models and keep the categories that suit the style.