Grok Imagine Image 2.0: Image Edit
Grok Imagine Image 2.0 performs precise image editing and text rendering using xAI's Aurora engine, letting you modify specific elements, swap backgrounds, or restyle scenes through natural language with crisp short-text output. It takes up to 3 reference images and generates at 1K or 2K resolution across 14 aspect ratios, with strong character consistency across scenes. Ideal for poster and cover design, product and brand visuals, or multi-scene character workflows.
试试一个示例
关于此工作流
This ComfyUI tutorial walks you through Grok Imagine Image 2.0: Image Edit, a prompt-based image editing workflow powered by xAI’s Aurora engine. The workflow is intentionally simple: LoadImage brings in your source and up to three optional reference images, GrokImageEditNodeV2 interprets your natural-language instructions to apply precise edits (like background swaps, restyling, or adding short text), and SaveImage writes the final render to disk. You can target 1K or 2K output and pick from 14 aspect ratios, making it easy to generate poster-ready or platform-specific visuals.
Under the hood, GrokImageEditNodeV2 conditions the edit on the provided base image plus any reference images to maintain visual identity and style. This enables strong character consistency across scenes, reliable product look alignment, and crisp short-text rendering for cover titles or label callouts—without manual masking. Because the node centralizes key settings (prompt, aspect ratio, and resolution), you can iterate quickly: refine the prompt, swap a reference, and re-render until the composition and typography meet your needs.
常见问答
You can connect up to three reference images to GrokImageEditNodeV2. Use them to preserve subject identity, transfer style, or anchor product/brand details so the edit remains consistent across scenes.
No mask is required. GrokImageEditNodeV2 applies edits from natural-language instructions. For tighter control, use precise prompts (e.g., describe the subject and the background separately) and start from a well-framed base image.
The node renders at 1K or 2K resolution and offers 14 aspect ratios. Choose the aspect ratio that matches your output target (e.g., poster, square, vertical) to avoid unwanted cropping later.
Keep text short (titles, labels, or brief slogans), specify the exact string, case, and placement in the prompt, and request high contrast against the background. If needed, generate a clean background or reserved space first, then iterate on type styling.