Nano Banana 2.1 covers both workflows on one model: images from a description, or edits driven by up to ten of your own reference images, at 1K, 2K or 4K.
One selector, two modes. Text to image starts from a description; image to image starts from at least one uploaded reference. Both share the same quality tiers and the same shape list.
The shape list is unusually wide, with extreme banner ratios alongside the standard ones, which is useful when the destination is a fixed slot such as a header or a story frame.
Choose a quality tier and a shape, or leave the shape on automatic. In editing mode, upload the references you want the model to work from; file output is PNG or JPG.
Switch between generating from text and editing from references without changing models or relearning controls.
Supply a set of images to keep a subject, product or style consistent across the task.
Standard ratios plus very wide and very tall banners, including an automatic setting.
Keep a product or character consistent while varying the scene across several images.
Use the extreme ratios for fixed slots instead of cropping a square render afterwards.
A general-purpose default when a task does not need background removal or a specialist tool.
Use a button on this page. It opens the right workspace with this model already selected, in the language you are reading now.
Write what should happen in the shot or image. Concrete subjects, framing and action give the model more to work with than short labels.
Choose length or quality tier, shape and sound where offered, then read the credit quote for those exact settings before you generate.
Generation uses account credits. Cost depends on the model, quality and output size or duration. Reference images, video or audio may add charges in some modes. Check the workspace quote for your selected inputs and settings before generating; the pricing page lists credit packages.
See credit packagesWe do not publish capability comparisons between versions. On this platform Nano Banana 2.1 offers three quality tiers, up to ten references and a wide shape list; check both models in the workspace selector and compare on your own prompt.
Not as a selectable option. For background work use a model with an explicit background control, or a dedicated background tool.
Up to ten images in one editing task.
PNG and JPG. Pick the format in the workspace before generating.
Other models on this platform that do a similar job, or that pair well with this one.
The FLUX 3 family product available on this platform: text generation and reference editing at 1K, 2K or 4K, with PNG, JPEG and WebP output.
Read the guideOne of the two selectable GPT Image 2.5 products on this platform: text and reference-based image generation with transparent-background support at the 1K tier.
Read the guideWan 3.0 is the standard variant of the Wan 3.0 family on this platform: a wide length range, three quality tiers and shape control in both text and image mode.
Read the guide