Grok Imagine Image 2.0

Grok is xAI's image model - the one people kept pointing at for photorealistic people and for being the least restrictive mainstream generator around. Until recently, "use Grok" meant going to X/Twitter and fighting the chat UI. This node drops Grok Imagine Image 2.0 into your ComfyUI graph as a plain generator node, running on Runware's cloud. Closed weights, no local option - this is exactly the case where an API node is the right tool rather than a compromise: you were never going to download this model, so paying per call is the only way to get it next to your local upscaler.

Image 2.0 is xAI's current image generation and editing model, released through Runware in August 2026. Per Runware's model page it keeps the family's core workflow - aspect-ratio control, image-based editing - and adds a quality parameter so you can tune output fidelity. If you've used Grok on X, the behavior carries over; the community read on Grok has long been strong photorealism of people plus a looser content line than Google or OpenAI enforce.

What you set. The one required input is positivePrompt. Everything else is optional, and the ones that earn their keep:

• referenceImages - this is how you do image-to-image and keep a subject or style consistent. Wire in an IMAGE and the model transforms or matches it rather than generating from nothing. For an API without seeds-and-LoRA workflows, references are your main lever.
• size - 29 presets, defaulting to 1K (16:9), with options like "1K (from input)" / "2K (from input)" for editing jobs and phone-ish ratios (9:19.5, 9:20). Pick one rather than fighting for a resolution widget that doesn't exist here.
• settings.quality - low or medium; the node only exposes those two. Start at the default medium, drop to low when you're iterating on prompts and want cheaper drafts.
• numberResults (up to 20) and the watermark/safety.checkContent group are self-explanatory if you've used hosted image APIs before.

Output is a native IMAGE called image, so wire it into Preview Image or Save Image and it looks like any sampler output - even though the generation happened on xAI's backend through Runware.

Install. One pack covers every Runware model, so this is the shared setup:

# ComfyUI Manager (recommended): Custom Nodes Manager → search "Runware" → install → restart
cd ComfyUI/custom_nodes
git clone https://github.com/Runware/ComfyUI-Runware   # manual alternative
pip install -r ComfyUI-Runware/requirements.txt

Restart, then drop in a Runware API key (Settings → Runware API key, or RUNWARE_API_KEY). No model downloads - it's all cloud.

Gotchas. Three, all honest API-node realities. Every queue is metered, and the node prints the cost in its title bar - Grok is a popular model, not a cheap one, so check that number before you queue twenty drafts. Your prompts and reference images leave the machine. And the content line belongs to xAI, not to you: the filter follows the model, and there's no local anything to patch, no matter what the Grok-on-X reputation suggests.