Image Models
Banana Flow ships hosted access to multiple state-of-the-art image generators. All of them bill against the same credit balance — switch models freely without juggling separate accounts or API keys.
Quick reference
| Model | Provider | Best for | Credits |
|---|---|---|---|
| Nano Banana 2 | Fast default, T2I + I2I, ultra-wide ratios | 8 | |
| Nano Banana Pro | Higher fidelity, longer prompt context | varies | |
| GPT Image 2 | OpenAI | Strong prompt adherence, ultra-wide 21:9 | 5 |
| Imagen 4 Ultra | Flagship realism | varies |
Credit costs are listed per generation. The actual amount is shown on the Output node before you click Generate.
Nano Banana 2
Powered by Google’s Gemini 3.1 Flash. Combines fast generation with pro-level image quality. Supports both text-to-image and image-to-image editing. Three quality tiers available — Standard (1K), High (2K), Ultra (4K).
Aspect ratios: square, portrait, landscape, plus ultra-wide options including 4:1, 8:1, and 21:9.
Pick this when: you want a strong default. Most new users start here.
Nano Banana Pro
Higher-fidelity sibling of Nano Banana 2. Longer prompt context window and better adherence on complex compositions. Trades speed for quality.
GPT Image 2
OpenAI’s latest image model. Long prompt context, very strong prompt adherence — useful when you need the model to follow a specific multi-part spec faithfully. Supports flexible aspect ratios including ultra-wide 21:9.
Pick this when: your prompt is long, structured, or carries lots of constraints.
Imagen 4 Ultra
Google’s flagship image model. Strong on photoreal scenes, materials, and lighting.
Choosing a model
- Need it fast and cheap? → Nano Banana 2
- Long, complex prompt? → GPT Image 2
- Maximum fidelity? → Nano Banana Pro or Imagen 4 Ultra
- Editing an existing image? → Nano Banana 2 (image-to-image)
The Output node lets you change models without losing prior generations — try the same prompt across two or three side-by-side and compare.