Nano Banana Image to Image

Nano Banana image to image gives you two models in the same workspace. Nano Banana 2 is a useful starting point when you want several attempts at a new setting, a different illustration treatment, or a change in lighting. Nano Banana Pro gives you another model to try against the same reference and brief. Its higher credit cost does not guarantee a better result for every photograph or instruction.

Decide what would make an edit successful before choosing between them. For a product scene, that might mean a readable label, a believable contact shadow, and the correct proportions. For a portrait, it might mean a convincing expression and a background that fits the light. Compare those specific details across results rather than judging only the overall style of a small preview.

Nano Banana 2 currently costs 5 credits per image, and Nano Banana Pro costs 10. Both offer 1K, 2K, and 4K output in this editor. The selected resolution currently leaves their credit cost unchanged. Each request produces one output, so a second attempt is a separate generation. Check the cost beside the generation button when you change models.

A new account receives 10 one-time signup credits, enough for 2 Nano Banana 2 attempts or 1 Pro attempt at current rates. The gift credits do not expire. Sign in and complete email verification before generating. The free image to image guide explains the starting allowance and what happens when you need more credits.

Choose your aspect ratio for the intended placement before submitting. A wide scene and a vertical cover need different framing, even with the same subject. Higher resolution provides a larger output; it does not establish that an inaccurate object or face has been corrected.

Begin with a photograph whose main subject is easy to distinguish from its surroundings. Blurred edges, heavy compression, and reflections across important lettering can make the desired edit harder to assess. Use an original file when available. This workspace accepts JPG, PNG, and WebP files up to 10 MB each, with up to five reference images.

Give additional references a clear purpose: another angle can explain an object, while a separate image can suggest a palette or setting. Describe which reference supplies each part of the brief. Conflicting views and several unrelated visual styles create more decisions for the model. You can start with one reference and add another only when a particular detail needs explanation.

Separate the features you want to retain from the change you want to explore. Identify the subject, its placement, and any important colors first. Then explain the replacement environment, the direction of the light, or the visual treatment. Concrete relationships, such as where a shadow falls or how much space surrounds the object, are easier to review than a long list of mood words.

This is whole-image editing through text instructions. The workspace does not provide a brush or mask to isolate a selected region. Even a request to change only the background can affect the subject. Download promising results, compare them with the original, and use the best candidate as a new reference when another pass is useful. Return to the original if successive edits begin to drift.

Inspect edges, hands, facial features, logos, lettering, and repeated patterns at a readable size. Check whether the perspective of the new scene agrees with the subject and whether the lighting direction is consistent. Reference images guide generation; they do not lock identity, exact text, or every original pixel. Keep your source image so you can check changes that are easy to miss in isolation.

If separate quality settings would help you plan a budget, consider GPT Image 2. It exposes Standard, Medium, and High quality alongside resolution, with costs varying by both settings. Compare an equivalent brief and review the actual outputs. The image to image model comparison collects the available controls and credit prices.