excel_Qwen_camera

Camera-move prompts are one of those things that's easy to get almost-right and annoying to phrase consistently - "slight tilt down" versus "looking down slightly" versus "downward angle" can all mean different things to a model, and remembering your own phrasing across a dozen generations is tedious. excel_Qwen_camera sidesteps that: instead of typing camera direction, angle, lens, and view-type language by hand, you pick each one from a dropdown and it assembles a single prompt string for you.

Despite living in the prompt/excel family by naming convention, this node has nothing to do with spreadsheets - it doesn't touch a file at all. It's purely a structured prompt builder, filed here alongside the pack's actual excel tools. Given the "Qwen" in the name and that Apt_Preset's own README documents a dedicated "QwenEdit mode" loader combination, this reads as a companion tool for Qwen-Image-Edit-style camera-control and multi-view workflows specifically, though the schema itself doesn't state that explicitly.

How it works

Five dropdown fields cover five independent axes of camera language, and every one of them defaults to "None" - leave whichever axes you don't care about untouched and only set the one or two that matter for a given shot:

• Pan direction - forward, backward, left, right, or one of the four diagonals.
• Angle adjustment - horizontal rotation left/right, tilted rotation left/right, looking down or up, forward/backward tilt, or clockwise/counterclockwise rotation.
• Angle value (0–180, step 5) - a numeric degree to pair with whichever angle adjustment you picked.
• Lens type - wide angle, ultra-wide, overhead, low-angle, close-up, extreme close-up, macro, near/medium/long shot, or panoramic.
• View type - a complete 4-view or 6-view template, or a single front/side/back/top/bottom/45°-half-side/30°-side view.

It combines whichever fields you set into a single prompt string. All five fields display their choices in Chinese in the node UI, with no separate English labels visible - worth knowing going in if you don't read Chinese, since you'll be pattern-matching against a reference (or the translations above) rather than reading the dropdown directly.

The inputs and outputs that matter

All five inputs are required but every one defaults to "None," so you only need to touch the ones relevant to your shot. There's a single output: the assembled prompt string, ready to feed straight into your text encoder or wherever your workflow takes prompt text.

How to install it

Bundled with the full pack. Via ComfyUI Manager: search "ComfyUI-Apt_Preset" → Install → restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/cardenluo/ComfyUI-Apt_Preset.git

Windows: run install.bat from the cloned folder through ComfyUI's embedded Python. Linux/Mac: activate your ComfyUI venv and install requirements from that folder by hand. Restart ComfyUI. This node is pure string assembly with no model files or heavy dependencies of its own.

Common issues & troubleshooting

Dropdown labels are in Chinese and you don't read Chinese. This is real friction and there's no built-in localization toggle visible in the node. Keep a translation reference handy (the field groupings above map roughly to pan direction / rotation-and-tilt / degree value / lens type / multi-view template) until you've memorized the layout.

Output prompt has less effect than expected. Remember this node only produces camera-framing language - it says nothing about subject, style, or lighting. It's meant to be combined with the rest of your prompt, not to replace it; concatenate its output with your subject description rather than using it alone.

Setting both a pan direction and an angle adjustment produces a confusing combined instruction. These two axes aren't mutually exclusive at the schema level, so nothing stops you from combining, say, "pan left" with "tilt down" - that's a legitimate compound camera move, but it's on you to make sure the combination makes physical sense before you rely on the model to interpret it correctly.