ArchAi3D_Qwen_Object_Position_Control

Product photography runs on one boring, reliable move: the camera gets closer or farther. ArchAi3D_Qwen_Object_Position_Control is that move, turned into a Qwen-Edit prompt. It's the smallest-scope camera node in the ArchAi3d Qwen pack - no orbits, no angles, no walkthroughs - and that narrowness is exactly why it's dependable.

The default preset is ecommerce_standard, which the author describes as the professional product-listing distance. So the out-of-the-box behavior is: give it a product description, and it produces a "move the camera closer to the product" instruction that reads like a real photographer's dolly move, plus a system prompt telling the model to keep the product, its surroundings and its identity fixed while the framing tightens or widens.

How it works

The core is the "dolly" prompt pattern, which the pack's QWEN_PROMPT_GUIDE rates as the most consistent zoom-family phrasing. prompt_style lets you pick between that verified dolly pattern and zoom (a user-tested alternative) - both are exposed because different Qwen revisions respond to one or the other slightly better. movement is simply in (closer) or out (farther), and speed adds a quality modifier - smooth for standard, cinematic for a fancier read, none to skip the modifier entirely when it's muddying a simple edit.

The inputs that matter

• movement - in or out. This is the whole show. In for detail shots, out for context.
• custom_target - what to push toward: "the watch face", "the logo", "the handle". Defaults to "the product" and gets you a lot further if you name a specific part.
• scene_context - a product description like "stainless steel watch" or "coffee mug on wooden table." Improves consistency more than any other field.
• preset - a few professional framing presets if you don't want to fiddle; ecommerce_standard is the default for good reason.

Outputs follow the pack standard: prompt and system_prompt (the useful ones - wire to your Qwen encoder's user-text and system-prompt inputs), plus position_description describing the preset you picked.

Installing it

Ships inside ComfyUI-ArchAi3d-Qwen by Amir Ferdos (ArchAi3d). Install once, use all the pack's nodes:

# ComfyUI Manager: search "ArchAi3d Qwen"
cd ComfyUI/custom_nodes
git clone https://github.com/amir84ferdos/ComfyUI-ArchAi3d-Qwen.git
cd ComfyUI-ArchAi3d-Qwen
pip install -r requirements.txt

Restart ComfyUI. The node itself is just a text generator - you bring the Qwen-Image-Edit model (GGUF-quantized in practice) and the encoder to run it.

Common issues

• The product gets remodeled. Qwen-Edit regenerates the whole frame, so "dolly in" can become "redesign the product." Keep the system prompt connected - it carries the "keep everything identical" language that holds the product together.
• Distance isn't a slider. This node phrases movement in meters via the pattern, but Qwen's reading is approximate. Direction is consistent; exact framing is not.
• Off-center subjects suffer. The pack's research notes centered subjects work best - if your product is parked at frame edge, crop-first or expect the model to drift it toward center.

When you just want "closer" or "farther" without the orbit machinery, this is the node. It won't do 360s - but it's the one you'll reach for most often anyway.