Qwen Image 3.0 (BudgetPixel)

Qwen Image 3.0 is the node that should probably be your default Qwen. Alibaba's own framing is "unified generation and editing balancing quality and speed," and the price makes that concrete: 40 credits per image at 1K, 45 at 2K. That undercuts Qwen Image 2.0 (45 credits, 1.5K default) for the same generation-and-editing feature set, with a newer engine on top. When the newer model is also the cheaper one, the decision makes itself - the only reason to go with the 2.x line is a workflow pinned to its exact behavior.

It's a hosted API node, like the rest of the BudgetPixel pack: prompt, negative prompt and up to three reference images go to the API, the render happens server-side, and the image comes back as a core IMAGE tensor. No local GPU, no model download, and no local option for this Qwen at all - the endpoint is closed-weight here.

The inputs

• prompt (required) - description or edit instruction.
• negative_prompt - what to avoid.
• reference_images - up to three, for editing and multi-image fusion. Omit for text-to-image; match_input_image keeps the input's aspect ratio.
• size - 1K (default, 40 credits) or 2K (45 credits). The price is per image, so four images at 1K is 160 credits.
• aspect_ratio - nine options.
• seed - -1 for random, or set for reproducibility.
• num_images - 1–4.

Output is images into Preview Image or Save Image.

The honest positioning

If you want a Qwen node with editing, a negative prompt, and multi-image fusion - the full modern feature set - 3.0 is the value pick. It doesn't have the "fine-grained realism" push of 2.0 Pro (85 credits), so for photorealistic client work Pro still earns its premium. And it doesn't have base Qwen-Image's rock-bottom 15-credit typography special, though its text rendering is perfectly good. But for the everyday "generate, edit, don't overpay" middle of your workflow, 3.0 is the sweet spot, and at 40 credits it's only 25 more than the pure text-to-image base node.

The pack's Qwen ladder in one line, cheapest to priciest: Qwen-Image (15, text-to-image), Qwen Image 3.0 (40–45, balanced gen+edit), Qwen Image 2.0 (45, fast gen+edit with negative prompt), Qwen Image 2.0 Pro (85, premium realism). If you're starting fresh, start here.

Install and issues

Install via ComfyUI Manager (search "BudgetPixel") or:

cd ComfyUI/custom_nodes
git clone https://github.com/budgetpixel-ai/comfyui-budgetpixel.git
pip install -r comfyui-budgetpixel/requirements.txt

Restart, find it under Add Node → BudgetPixel → Image, and set your API key via BUDGETPIXEL_API_KEY or a budgetpixel.json file - no key widget, so shared workflows stay safe to share.

Pack-wide caveats: a recent ComfyUI for core types, and interrupting a run abandons the local wait while the server-side job keeps running and bills its credits - low stakes at 40, but it's the pack's rule. And as a hosted model, the content filter is whatever Alibaba's endpoint enforces; a negative prompt works within those walls, not around them.