Simple Angles Selector
Three dropdowns, one camera prompt — the boring one you'll actually script
- prompt
This is the no-frills sibling in the ComfyUI-AnglesSelect pack, and it might be the one you actually want. Where AnglesSelector3D gives you a spinning Three.js sphere full of clickable dots, SimpleAnglesSelector is just three dropdowns: pick an azimuth, an elevation, and a distance, and out comes one prompt string ready for the fal Multiple-Angles LoRA on Qwen-Image-Edit-2511. No widget, no cat, no fuss.
If you've ever fought a prompt generator, you'll appreciate the honesty here. The LoRA expects a strict vocabulary - <sks> trigger token followed by a camera phrase in a specific order - and this node makes it impossible to get that format wrong. Every option it offers maps to exactly one of the phrases the LoRA was trained on, so you can't fat-finger "from above looking down a bit" and get nothing back.
How it works
It's a plain mapping node. Each dropdown label you pick gets converted to a number (azimuth in degrees, elevation in degrees, distance as an index), and those numbers get translated back into the LoRA's English camera phrases. The dropdown says "Front-Right"; the prompt says front-right quarter view. The labels themselves switch language automatically based on your system locale - a zh locale gets 方位角/仰角/距离 with Chinese options, everything else gets English.
The inputs and output
There are only three inputs, and they're all dropdowns:
- Azimuth - 8 options: Front, Front-Right, Right, Back-Right, Back, Back-Left, Left, Front-Left. Defaults to Front.
- Elevation - 4 options: Low Angle, Eye Level, Elevated, High Angle. Defaults to Eye Level.
- Distance - 3 options: Close-up, Medium, Wide. Defaults to Medium.
The output is a single prompt STRING, something like:
<sks> front-right quarter view elevated shot wide
In the pack's example workflow that string feeds straight into a TextEncodeQwenImageEditPlus node as your positive conditioning, then on to the KSampler. The LoRA itself - and the Qwen-Image-Edit UNET, a couple of LoraLoaderModelOnly nodes, and the image you're editing - are all wired up around it.
Installing it
Same as its sibling, because it's the same pack: ComfyUI Manager (search "ComfyUI-AnglesSelect"), or
cd ComfyUI/custom_nodes
git clone https://github.com/jianglinbin/ComfyUI-AnglesSelect
then restart. There's no requirements.txt and no model download - nothing to configure. The whole repo is two node classes plus the bundled Three.js assets.
Why you'd pick this over the 3D node
Two reasons. First, when you just want one angle, the dropdown is faster than orbiting a sphere. Second, and this is the one that matters: a node with three plain enum inputs is trivially scriptable. The community's whole multi-angle play runs the edit workflow hundreds of times (that's what made the Multiple-Angles LoRA famous), and for batching, looping, or driving the graph from the API, three dropdowns are infinitely easier to automate than a JSON widget you'd have to serialize by hand. Batch-happy people will find themselves reaching for this one.
It's a two-minute node, and it knows it. It doesn't try to be impressive - it just makes sure that when you tell Qwen-Image-Edit where the camera should go, you use the exact words it was trained on.
Inputs (3)
| Name | Type | Default | Description |
|---|---|---|---|
| Azimuth | COMBO | Front | 8 options: Front, Front-Right, Right, Back-Right, Back, Back-Left, +2 |
| Elevation | COMBO | Eye Level | 4 options: Low Angle, Eye Level, Elevated, High Angle |
| Distance | COMBO | Medium | 3 options: Close-up, Medium, Wide |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| prompt | STRING | — |