Image
Google

Nano Banana 2

Google DeepMind's fast image model for text-to-image and reference-guided editing at up to 4K.

À partir de 60 HGcoins / génération·paiement à la génération, sans abonnement
Généré avec Nano Banana 2
Généré avec Nano Banana 2
Exemples

Créé avec Nano Banana 2

Exemples de rendus. Ouvrez Studio pour générer les vôtres.

À quoi ça sert

Là où Nano Banana 2 excelle

Text to image

Generate original images from a written prompt at 1K, 2K, or 4K across 15 aspect ratios.

Reference editing

Add up to 10 reference images to guide image-to-image edits toward an existing subject or style.

Batch variations

Produce up to 8 prompt variations in one batch to compare directions quickly.

Points forts

  • Supports both text-to-image and reference-guided image-to-image editing in one model
  • Accepts up to 10 reference images to steer image-to-image generation
  • Selectable output resolution across 1K, 2K, and 4K
  • 15 aspect ratios from square and widescreen to extreme panoramic 8:1 and 1:8
  • Generates up to 8 prompt variations in a single batch
  • Built on Google DeepMind's Gemini 3.1 Flash Image for fast, lower-cost generation

Compromis

  • Image only, with no video or animation output
  • No audio output
  • Pricing increases at the 4K resolution tier with a surcharge over the base per-image cost
  • Image-to-image mode requires supplying at least one reference image
Specs

En un coup d'œil

Type
Image
Vendor
Google (DeepMind)
Modes
Text-to-image; image-to-image (with references)
Resolution
1K / 2K / 4K (default 1K)
Reference images
Supported, up to 10
Output
PNG

À propos de Nano Banana 2

Nano Banana 2 is Google DeepMind's image generation and editing model, also known as Gemini 3.1 Flash Image. Launched in February 2026, it is positioned by Google as a fast, lower-cost option that brings strong image quality and world knowledge at Gemini Flash speed. On HexGen you can run it for both text-to-image creation and reference-guided editing in a single model.

Start from a written prompt to generate an image, or add up to 10 reference images to switch into image-to-image mode and steer the result toward an existing look, subject, or composition. Output resolution is selectable across 1K, 2K, and 4K, and you can choose from 15 aspect ratios spanning square, widescreen 16:9 and 21:9, portrait 9:16, and extreme panoramic shapes like 8:1 and 1:8.

Need options fast? Nano Banana 2 can produce up to 8 prompt variations in a single batch, so you can compare directions in one pass. Every image is delivered as a PNG, ready to download from your HexGen workspace.

Idées de prompts

Points de départ

Copiez, ajustez et lancez. Un bon prompt fait l'essentiel du travail.

A cozy Scandinavian living room at golden hour, soft natural light through tall windows, warm wood floors, photorealistic, 16:9.

Studio product shot of a frosted glass perfume bottle on a marble surface, dramatic side lighting, clean background, 4:5.

Using the attached reference photo, restyle the portrait as a vintage 1970s film look with warm grain and muted colors.

Tarifs
60
HGcoins / génération · ≈ $0.06

Payez seulement ce que vous générez. 1 USD = 1,000 HGcoins. Les HGcoins n'expirent jamais et les échecs sont remboursés automatiquement.

Comparer

Nano Banana 2 face aux autres modèles

Nano Banana 2 is Google's fast image model for both generation and reference editing. Here is how it sits next to other image models in the HexGen catalog.

Nano Banana 2 face aux autres modèles
ModèleQualitéVitesseCoûtÀ choisir quand
Nano Banana 2
Celui-ci
Google
Très bon
Ultra-rapide
Coût réduit
Best when you want fast, lower-cost text-to-image and reference editing in one model, with 1K to 4K output.
Excellent
Rapide
Coût élevé
Step up to the Pro tier when top-end image quality matters more than speed or cost.
ByteDance
Excellent
Rapide
Coût moyen
A strong ByteDance alternative if you want to compare image quality across vendors.
En résumé : choisissez Nano Banana 2 quand best when you want fast, lower-cost text-to-image and reference editing in one model, with 1k to 4k output.. Sinon, l'un des modèles ci-dessus conviendra mieux. Touchez une ligne pour comparer.

Questions fréquentes

It generates images from text prompts and edits images using references. Both text-to-image and image-to-image modes are supported in a single model.