Qwen Image 2.0 Pro (BudgetPixel)

Qwen Image 2.0 Pro is the top of Alibaba's Qwen image ladder in this pack, and the pitch is in its own description: premium generation with enhanced text rendering, fine-grained realism, and stronger semantic adherence. What that adds up to, in practice, is an image that obeys the prompt more literally and looks more like a photograph rather than a rendering - the "client-grade" Qwen. It costs 85 credits per generation, nearly twice the standard 2.0, and you pay it for the times when a miss means a redo.

It's a hosted API node like everything else in the BudgetPixel pack. Prompt, negative prompt and up to three reference images go to the API, the heavier Pro pipeline renders server-side, and the image returns as a core IMAGE tensor in your graph. No local GPU, no model files, no weights to download - and also no way to run this Qwen locally, since the Pro endpoint is closed-weight here.

The inputs

• prompt (required) - description or edit instruction.
• negative_prompt - what to avoid; same field as the standard 2.0.
• reference_images - up to three, for multi-image fusion and editing. Omit for text-to-image; use match_input_image aspect ratio to keep the input's frame.
• size - 1K / 1.5K (default) / 2K.
• aspect_ratio - nine options.
• seed - -1 for random, or set for reproducible results.
• num_images - 1–4.

Output is images into Preview Image or Save Image.

Pro vs the rest of the Qwen family

The honest take: Pro is the same interface as Qwen Image 2.0 with a bigger engine behind it. The differences you'll actually notice are the "fine-grained realism" - skin, fabric, natural light behave more convincingly - and semantic adherence, meaning complex prompts with several conditions are obeyed instead of approximated. If your work is photorealistic product or portrait assets where the seams show, the 85 credits is a real upgrade over the 45-credit standard. If you're making stylized or illustrative images, Pro's realism emphasis buys you less; the standard 2.0 or the newer 3.0 (40 credits) is a better spend.

One thing worth knowing: this node has no num_images-by-resolution surprise - 1.5K default, up to 2K - so the price is per generation regardless of how many images you ask for. That's a subtle plus when you're batching four images in one call.

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, by design.

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 still bills its 85 credits - the most expensive "oops" in the Qwen family, so check the Cost Estimate node if you're unsure. And like all hosted models, the filter is baked in: negative_prompt steers within the vendor's rules, not around them.