Nano Banana 2 Lite: Image Edit
Edit an image using Nano Banana 2 Lite, the fastest image generation model from Google’s Gemini series, producing a side-by-side comparison of the original and edited result.
试试一个示例
关于此工作流
This comparison workflow shows a clean, side-by-side view of how Nano Banana 2 Lite (via the GeminiNanoBanana2V2 node) transforms a source image based on your edit instruction. The pipeline loads your original image, applies the edit with the model, and then places the original and result next to each other using ImageCompare so you can quickly judge what changed and what stayed consistent.
Methodologically, the comparison is controlled: both panes derive from the exact same source image (LoadImage), and only the model-driven edit differs. This makes it easy to evaluate trade-offs such as speed versus fidelity, the extent of stylistic drift, and the preservation of structure or identity. Expect Nano Banana 2 Lite to prioritize responsiveness and efficient edits; in practice, that can mean fast, coherent changes for common tasks (color tweaks, small object additions/removals), while very intricate or large, global transformations may require tighter instructions or iterative passes.
常见问答
The workflow compares the original image (from LoadImage) with the edited image generated by GeminiNanoBanana2V2. ImageCompare places them side-by-side for visual assessment; no numeric metrics are computed.
Keep the source image constant and change only the edit instruction or a single parameter at a time. Use consistent output size, and review the same regions (faces, edges, text) for drift or artifacts.
Choose it when you need fast, coherent edits to an existing image—such as color tweaks, lighting adjustments, or minor object-level changes—where turnaround speed is important.
Under-specified instructions can lead to broader changes. Make your intent explicit (what to change and what to keep), and iterate with more targeted guidance to reduce unintended drift.