Nano Banana: imágenes IA de Google desde texto o referencia
Carga de imágenes
Carga una imagen para usar como referencia, máximo 5 imágenes permitidas.
Las relaciones nativa, vertical, horizontal y cinematográfica ahora se renderizan directamente sin reprocesos ni créditos adicionales.Relaciones oficiales: 1:1 • 9:16 • 3:4 • 2:3 • 4:5 • 16:9 • 4:3 • 3:2 • 5:4 • 21:9
Galería de salida
Tus creaciones de IA ultrarrápidas aparecen aquí instantáneamente
Sobre Nano Banana
Nano Banana is Google's image model, and this page runs both of its modes: text to image, where a written idea becomes a finished picture, and image to image, where up to five reference images hold a subject, product or style steady while you change the scene around it. Whichever mode you pick, a run costs 30 credits.
Most image pages ask you to trade control for speed. This one leans toward volume without giving up the part of control that is hard to replace. A run costs 30 credits, an eighth of a Nano Banana Pro run, so a series of tests stays affordable. What you keep at that price is reference-image editing and native aspect ratios: the picture is composed for the frame you selected, not cropped into shape afterwards.
Text to image starts from the prompt alone. Image to image accepts one to five uploaded images in a single request, which is where this model separates itself from single-reference editors: hand it four angles of the same product and a background plate, then describe how they should combine. Uploads accept PNG, JPG, JPEG and WEBP at up to 10 MB each. Write the instruction as a change to what is already in the frame, because the model has the pixels and needs the intent.
This is not a page for art direction. There is no stylization dial, no version selector and no speed tier: one model, one price, both modes. When you want a look pushed toward illustration or a period reference, Midjourney exposes those dials; when you want a single photo rewritten while everything around the change stays put, Flux Kontext is the tighter tool. What this page does guarantee is the unglamorous part that matters at scale: failed jobs return their credits, and prompts can be translated to English before they are sent.
Reference count does not change the price. Whether an image-to-image request carries one upload or five, the run is still 30 credits, so the real limit on references is attention rather than budget. That has a practical consequence: a fifth angle that disagrees with the other four makes the set worse, not richer. Give the model references that agree on lighting, scale and viewpoint, and use the prompt for the one thing the set cannot show, such as the new surface or the new season.
The workspace also keeps your work through sign-in. If you write a prompt, pick a ratio and press generate without a session, the page sends you to log in with the mode, prompt, ratio and model carried in the address, and the editor comes back filled. The attempt itself costs nothing; billing starts only when a signed-in run is accepted by the model. That makes the page usable as a drafting surface before you have an account, provided you finish in the same browser and do not clear the tab.
Cómo funciona Nano Banana
- 1
Pick the mode
Text to image when you are starting from nothing; image to image when the subject already exists. A shared link can preset the mode, prompt and aspect ratio, so the workspace opens with the job half written.
- 2
Describe the change, not the scene
In image to image, name the edit and the part of the picture it applies to, such as swapping the surface a product stands on or restyling one garment. Everything you leave out of the prompt is carried by the reference images.
- 3
Set the frame, then generate
Choose from the native ratio list, from 1:1 through 21:9, so nothing is cropped later. Pick PNG or JPEG, confirm the 30-credit charge on the button, and run.
Para qué sirve Nano Banana
Product variants at volume
Upload five angles of one item, then change only the scene in each prompt. Because the reference set stays fixed, the product does not drift between images.
Character and mascot sets
Reuse the same reference images across poses and keep a face or costume recognisable while the situation changes around it.
One idea, every placement
Generate the same concept at 1:1 for feed and 9:16 for stories from the native ratio list, with no reframe delay or second crop.
Cheap concept sweeps
Thirty credits buys a rough direction. Test ten before committing, then move the winner to Nano Banana Pro for the final image.
Nano Banana de un vistazo
- Model
- Google Nano Banana, run through KIE
- Credit cost
- 30 per generation in both modes
- Modes
- Text to image, image to image
- Reference images
- Up to 5 per request, PNG/JPG/JPEG/WEBP, 10 MB each
- Aspect ratios
- Auto, 1:1, 3:4, 2:3, 4:5, 4:3, 3:2, 5:4, 16:9, 9:16, 21:9
- Output
- PNG by default, JPEG optional
Preguntas frecuentes sobre Nano Banana
- How much does one Nano Banana image cost?
- 30 credits per generation, in either mode. A free account collects 50 credits each day you log in and a 300-credit bonus at signup, so the first image can be generated the same day you register. Starter is $4.90 a month and includes 300 credits.
- Can I use several reference images at once?
- Yes, up to five in one image-to-image request. That is what makes consistent product and character sets possible: the same uploads can drive a whole series of scenes, and you only rewrite the part of the prompt that changes.
- Should I use this page or Nano Banana Pro?
- Pro is 240 credits per image and is built for the higher-fidelity pass. This page is for deciding what the image should be; move the chosen direction to Pro when you need the keeper. Both accept reference uploads and both use native aspect ratios.
- What happens when a generation fails?
- The 30 credits are returned automatically. Billing follows the delivered image, so a provider error or a job that never completes does not cost you anything.
- Which aspect ratios are supported?
- Auto plus ten fixed ratios: 1:1, 3:4, 2:3, 4:5, 4:3, 3:2, 5:4, 16:9, 9:16 and 21:9. They render natively rather than being cropped from a square, so a 21:9 banner keeps the composition you asked for.
- Do I lose my prompt if I have to sign in first?
- No. The sign-in link carries the mode, the prompt, the aspect ratio and the selected model, so the editor reopens with the same text and settings once you are back. The interrupted attempt does not consume credits, because billing only starts when a signed-in run is accepted. Finish in the same browser and keep the tab, since the settings travel in that link rather than in an account-side draft.