Simple Camera Selector

Here's a niche problem that's more common than it sounds: you want a vision model to caption an image, but you need the caption to describe the photography - the lens, the framing, the lighting - not just "a woman standing on a street." Simple Camera Selector exists for exactly that job. It's the camera-preset sibling of the pack's Style Selector, built for captioning tasks where photographic context matters.

The idea: for training data or prompt libraries, a caption that says "a woman on a street at golden hour, shot on a 85mm lens from a low angle, shallow depth of field" is worth ten times a plain description. This node makes that reproducible by appending a curated camera/lighting/framing block to your user prompt, pulled from a dropdown.

Inputs and outputs

• camera_preset - the dropdown. Starts with No changes and Random, then 30+ built-in camera presets with names like "Dreamy Golden-Hour Portrait", "Cinematic Telephoto Close-Up", "Candid Street Moment, Unposed". Random re-rolls on every execution, which is nice for building a varied captioning dataset in a batch.
• user_prompt - optional prefix, prepended before the camera text.

Outputs:

• user_prompt - your prefix + the camera preset text. Wire this into the main Qwen node's user_prompt input.
• camera_name - the applied preset's name as a string, so you can log or save which one was used.

Presets live in the _camera_preset section of system_prompts_user.json, same config-file pattern as the other selectors. Add your own camera descriptions there if the built-ins don't cover your style - e.g. a macro product shot, a drone aerial, a CCTV still.

Install

Same pack as the main node - KLL535/ComfyUI_Simple_Qwen3-VL-gguf. If you have that installed for Qwen captioning, this is already sitting in the 🌐 SimpleQwenVL category:

cd ComfyUI/custom_nodes
git clone https://github.com/KLL535/ComfyUI_Simple_Qwen3-VL-gguf

or via ComfyUI Manager (search ComfyUI_Simple_Qwen3-VL-gguf), then restart and F5.

When to reach for it

This is a specialized tool and it knows it. The README is honest: camera presets "can sometimes give interesting results," which is a diplomatic way of saying they work well for image-captioning pipelines and are overkill if you just want a quick "describe this" answer. Pair it with a Detailed Description system preset from Master Prompt Loader and you get the full photographic-caption recipe; on its own with a terse user prompt, the model may ignore the camera angle and just describe the scene. If you're building captioning training data, though, it's the difference between generic and actually useful alt-text - and it's a zero-dependency text node, so there's nothing to break. Pick a preset, wire the output to the LLM, done.