The biggest complaint among AI image generators is always the same: the character's face changes in the next frame, the outfit looks different. Google is tackling this problem with Nano Banana 2.1. The model also supports mask-based editing, letting you modify only the part of the image you select.
The pricing is bold too. Built on Gemini 3.6 Flash, the model offers 4K resolution at roughly half what competitors charge.
Price advantage stands out in the market
OpenAI's model currently leads image generation, with Microsoft's model released in August in second place. Google's trump card is cost.
The company cut standard API fees by 50 percent. One million image units dropped from $60 to $30. A 1K image costs $0.0336, while a 4K image comes in at $0.0756. That's an especially attractive proposition for developers.
Characters stay themselves, frame after frame
Nano Banana 2.1 can process 14 reference images in a single pass. Tracking four characters and ten objects simultaneously, the model preserves characters' identity and appearance through edits.
The goal: preventing the same character from becoming unrecognizable across different scenes.
Where will the model appear?
The new model won't be limited to the Gemini app — it will also power Google Search's AI mode and Google Ads. This move will broaden the model's reach.
Comments (0)
No comments yet. Be the first!