AI Photo Models

Mashup offers three image generation models, all based on Google's Gemini image families. Each has a different cost-quality tradeoff.

Nano Banana — 5 tokens per image

The everyday workhorse. Available on every paid plan. Fast (typically under twenty seconds per image), efficient with tokens, and produces high-quality results for the vast majority of product categories. Nano Banana is what we recommend you start with — generate four images on this model, see whether the results are good enough, and only step up if you need more.

Nano Banana Pro — 6 tokens per image

Google's flagship image generation model, used when you need the absolute best fidelity. Excels at:

• Small detail (text on packaging, logos, fine print).
• Fabric and material textures.
• Specular reflections (glass, polished metal, glossy plastics).
• Complex jewelry and accessories where the difference is in the millimetres.

Slightly slower than Nano Banana (typically thirty to forty-five seconds per image). Available on Pro and Ultra plans.

Nano Banana 2 — 8 tokens per image

Pro-quality output at Flash speeds. Use this when you want premium results in volume — for example, generating a complete sixteen-image set for a hero product. Faster than Nano Banana Pro per image, slightly more token-intensive, and the quality is essentially indistinguishable from Pro in most blind comparisons. Available on Pro and Ultra plans.

Which Model Should I Pick?

Default to Nano Banana for exploration. Step up to Nano Banana Pro when you have a final image in mind and want maximum fidelity, or to Nano Banana 2 when you are generating in volume and want a balance of speed and quality. You can mix and match within a single product — exploration with Nano Banana, final hero shot with Nano Banana Pro.