Nano Banana 2
High-speed image generation optimized for rapid iteration. Supports the most reference images of any model.
Capabilities
| Feature | Support |
|---|---|
| Text-to-Image | Yes |
| Image-to-Image (editing) | Yes |
| Max Resolution | 4K (4096 x 4096) |
| Reference Images | Up to 14 (10 objects + 4 characters) |
| Aspect Ratios | 1:1, 1:4, 1:8, 2:3, 3:2, 3:4, 4:1, 4:3, 4:5, 5:4, 8:1, 9:16, 16:9, 21:9 |
| Quality Tiers | 0.5K, 1K, 2K, 4K |
| Multi-turn editing | Yes (via thoughtSignature) |
| Negative Prompt | No |
| Inpainting / Mask | No |
Reference Images
Nano Banana 2 supports up to 14 reference images in a single generation:
- Up to 10 object images — The model will faithfully include these objects in the final image
- Up to 4 character images — The model maintains character consistency across the generation
Simply connect multiple upstream nodes to your Nano Banana 2 node, and the model will incorporate all reference images as composition guidance.
0.5K Quality (Draft Mode)
Nano Banana 2 uniquely supports 0.5K (512px) output — a fast draft mode for:
- Quick concept exploration
- Thumbnail previews before committing to high-res
- Batch generation where speed matters more than detail
Switch to 1K/2K/4K when you're ready for the final output.
Multi-turn Editing
Same as Nano Banana Pro — the model maintains thinking context across edits when used within the Gemini model family. See Nano Banana Pro: Multi-turn Editing.
Prompting Tips
- Use many reference images for complex compositions. Nano Banana 2 handles up to 14 references, making it ideal for mood boards and multi-subject scenes.
- Use @-mentions to direct each reference. With 14 images,
@mentions help the model understand which image is for what:"use @[Image 1] as the main subject, @[Image 2] for color palette, @[Image 3] for the background". See Referencing Images in Prompts. - Start at 0.5K for exploration, then upscale. Generate many variants quickly, pick the best, then regenerate at 4K.
- Quoted text renders well. The model can generate legible, stylized text for infographics, menus, and marketing assets.
Limitations
- No negative prompt support
- No inpainting/mask editing
- Preview model — API may change