Create a fresh image, revise a reference, or build a visual with readable text while iterating quickly. Nano Banana 2.1—also known as Gemini 3.1 Flash Image—is available in CawCut’s [Image Generation] node for both Text to Image and Image to Image.
1. Nano Banana 2.1 Strengths
| Strength | Application |
|---|---|
| Fast generation and conversational editing | Explore visual directions and refine a source image through focused follow-up instructions. |
| Strong prompt adherence | Describe the composition, subject, and details you need, and keep the request organized by priority. |
| Character and subject consistency | Use clear references and explicitly name the identity or design details that should carry through. |
| Improved text rendering | Create graphics with text by supplying exact copy, hierarchy, and placement. |
| High-resolution output choices | Choose from the available 1K, 2K, or 4K resolution options for the intended use. |
2. Modes and Inputs
Nano Banana 2.1 supports Text to Image and Image to Image in CawCut. In Image to Image, CawCut lists 1–14 reference images.
| Mode | Use it for | Typical input |
|---|---|---|
| [Text to Image] | Creating a new image from a written brief. | Prompt describing subject, composition, style, and exact text if needed. |
| [Image to Image] | Editing or transforming an existing image. | Source image or reference images and a focused edit instruction. |
Nano Banana 2.1 supports 1–14 reference images in Image to Image mode. CawCut lists Auto and multiple fixed aspect ratios, 1K, 2K, and 4K resolution, Minimal/Medium/High Thinking, and 1–4 images per run. See Image Generation for the available controls.
3. Prepare Prompts and References
For a new image, describe the subject, composition, setting, and visual style. For an edit, identify the source and state the requested change before listing what should stay unchanged.
What to avoid: Avoid vague copy instructions, conflicting references, and prompts that ask for unrelated edits in one pass. When text accuracy matters, provide exact wording and describe its position and visual hierarchy; inspect the result before using it as finished artwork.
When you provide multiple images, assign each a clear purpose, such as identity, pose, product details, palette, or layout. For information on passing images through connected workflows, see Image and Image Group Nodes.
4. CawCut Workflow and Settings
- Add an Image Generation node to a workflow.
- Choose [Text to Image] or [Image to Image].
- Select Nano Banana 2.1 and add references if needed.
- Write a prompt with the desired result and the details to preserve.
- Choose the aspect ratio, resolution, Thinking level, and number of images.
- Generate and review spelling, composition, subject consistency, and details.
Use a focused follow-up edit to refine a close result instead of reworking all of the instructions. Review the credit estimate before generating several variants; see Credits.
5. Limitations and Refinement
Text and fine visual details can still vary, and edits may change areas beyond the intended target. Clear references and focused instructions help, but inspect spelling, layout, and preserved details before delivery.
| If you notice | Try this |
|---|---|
| A subject detail changes unexpectedly | Provide a clearer source reference and name the feature to preserve. |
| Text is incorrect or misplaced | Supply exact copy and specify its position and hierarchy. |
| Too much of the image changes during an edit | Ask for one localized change and explicitly identify what must remain untouched. |
| The composition does not match the intended use | Set the aspect ratio and describe the layout before adding decorative details. |
Inspect the final image at its intended size, especially small text and fine details. For more model options, see Supported Models.