Nano Banana Pro
Premium image generation powered by Gemini 3 Pro. Best for production-quality outputs with professional camera simulation.
Capabilities
| Feature | Support |
|---|---|
| Text-to-Image | Yes |
| Image-to-Image (editing) | Yes |
| Max Resolution | 4K (6336 x 5504) |
| Reference Images | Up to 6 (objects + 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 | 1K, 2K, 4K |
| Multi-turn editing | Yes (via thoughtSignature) |
| Negative Prompt | No |
| Inpainting / Mask | No |
Multi-turn Editing
Nano Banana Pro is a thinking model that maintains reasoning context across edits. When you generate an image and then refine it (e.g., "change the background to blue"), the model remembers its previous composition decisions and makes targeted changes instead of starting from scratch.
This works automatically when:
- The upstream image was generated by a Gemini image model
- The current node also uses a Gemini image model
When you switch to a non-Gemini model, the editing context is discarded (different models can't share thinking context).
Prompting Tips
- Describe scenes, not keywords. The model's strength is deep language understanding.
- Be specific about lighting and atmosphere. "Golden hour side lighting with warm tones" works better than "good lighting."
- Reference images guide composition. The model treats all input images as peer references, including the "init image" in edit mode.
- Use @-mentions for precision. Type
@to reference a specific image inline:"make the subject match @[Image 1] but in the style of @[Image 2]". The model sees each image at the exact position you mention it. See Referencing Images in Prompts.
Limitations
- No negative prompt support (describe what you want, not what to avoid)
- No inpainting/mask editing
- Slower than Nano Banana 2 (~45s vs ~15s)
- Maximum 6 reference images (Nano Banana 2 supports 14)