Grok Imagine 2.0 (BudgetPixel)

Grok Imagine 2.0 is xAI's "put it in the portfolio" model. Where the first Grok Imagine was about aesthetics, 2.0 is marketed around precision and range: exact editing, infographics, ads, game assets, UI/UX mockups, storyboards. That's a noticeably different pitch - it's positioning itself as the graphics-workhorse tier rather than the vibe tier. On BudgetPixel's shelf it costs more than the original (a quality × resolution matrix starting at 50 credits, versus flat 35), and the value proposition is that you're paying for control.

One detail worth flagging up front because it changes the math: when you supply an input image, the input image is free - you're billed for the output edit, not the reference. That's not how every node in this pack works, and it makes iterative editing much less painful than you'd fear.

How it works

Same BudgetPixel image-node flow - POST /images/grok-imagine-image-2, poll, download the IMAGE batch. The capabilities mirror the marketing: text-to-image from scratch, or single-image editing where the prompt describes the change ("turn this into a dark-mode version, keep the layout"). Note the input is a single image, not a multi-reference slot like GPT-Image-2 or FLUX 2 Dev - this model's edit path is one-input, instruction-driven, and it's genuinely good at following precise instructions, which is the whole reason you'd choose it over the cheaper original.

The inputs that matter

• prompt - required, multiline. With image attached this is the edit instruction; be specific, this model rewards it.
• image - optional. The single input image for editing. It's free, so don't feel guilty feeding it.
• quality - low / medium. Billing ladder: low 1K = 50, low 2K = 80, medium 1K = 80, medium 2K = 100 credits. Medium/1K is the sane default; the 2K steps are for when you need the pixels.
• size - 1K or 2K.
• aspect_ratio and num_images (1–4) round it out, plus seed.

Output is images (IMAGE batch).

Install

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, set BUDGETPIXEL_API_KEY or budgetpixel.json ({"api_key": "bpx_live_..."}) in the pack folder. No model downloads; requests is the only dependency.

Where people get burned

The billing tiers are the thing to read before you run: the tooltip spells out the quality × resolution grid, and it's easy to hit 100 credits (medium/2K) when 50 (low/1K) would do - run Cost Estimate with your exact {"quality", "size"} combo if you're unsure. Second, "precise editing" is still closed-model editing: xAI's moderation is lighter than OpenAI's but it exists, and the model refuses some content regardless of how precisely you phrase it. Third, don't confuse single-input editing with multi-reference composition - if your job needs four reference images stitched together, this node can't do it; that's GPT-Image-2 or FLUX 2 Dev territory.