Nano Banana 2 AI Image Generator Create, edit, and iterate with Gemini 3.1 Flash Image
Start from a written brief or upload reference images, then choose the format that fits the job. This Lovart ME workflow supports up to 14 references, 15 aspect ratios, 1K to 4K output, and JPG or PNG files.
- 14
- Reference images
- JPEG, PNG, or WebP inputs
- 15
- Aspect ratios
- From 1:8 to 8:1, plus Auto
- 4K
- Maximum output
- Choose 1K, 2K, or 4K
- 2
- Output formats
- Download as JPG or PNG
What Nano Banana 2 is useful for
Google positions Nano Banana 2 as Gemini 3.1 Flash Image: a fast image model for generation and editing. The practical advantage is a shorter path from a rough direction to a controlled, reusable visual.
Text-to-image drafts
Describe the subject, setting, composition, lighting, and required text in one brief. Use 1K while exploring, then raise the resolution after the direction is stable.
Reference-led editing
Upload one image for a focused edit or combine several references for subject, object, palette, texture, and layout cues. State which property comes from each reference.
Readable text and localization
Create posters, labels, cards, and mockups that include text. Put the exact wording in quotation marks and request a clear hierarchy; always proofread the result before publishing.
Consistent subjects and objects
Google reports consistency for up to five characters and fidelity for up to 14 objects in a workflow. Stable names, clothing notes, and reference order make repeat scenes easier to direct.
Format-aware composition
Choose the final channel before generating. A vertical story, square product tile, wide banner, and panoramic backdrop need different subject placement and negative space.
Choose resolution and aspect ratio before the final render
Higher resolution is useful when the image will be cropped, printed, or reviewed closely, but it also uses more credits. Set the composition first so the final render is not spent fixing a layout problem.
1K for exploration
Use 1K for prompt testing, composition choices, and early stakeholder review. It is the lowest-cost way to compare directions.
2K for everyday publishing
Choose 2K for web headers, product pages, presentations, and social assets that need more detail without the highest output cost.
4K for final detail
Use 4K when the approved image needs cropping room, larger displays, or print preparation. Inspect faces, hands, logos, and small text at full size.
A practical ratio map
Square and product
1:1, 4:5, and 5:4 suit catalog tiles, profile posts, and balanced product compositions.
Portrait and mobile
2:3, 3:4, and 9:16 leave room for people, covers, posters, and full-screen mobile stories.
Landscape and campaign
3:2, 4:3, 16:9, and 21:9 fit editorial images, presentations, banners, and cinematic scenes.
Extreme formats
1:4, 4:1, 1:8, and 8:1 are for narrow placements. Mention where the subject and copy area should sit so the composition does not collapse.
Nano Banana 2 prompt templates built around real deliverables
Replace the bracketed fields, keep the business requirement near the start, and remove any detail that does not affect the final image. These templates work as starting points, not magic phrases.
Product launch image
Use for a clean product visual with controlled space for campaign copy.
Create a [ratio] launch image for [product]. Keep the product shape and label from reference 1. Place it on [surface] with [lighting]. Leave clean negative space on the [side] for a headline. Brand colors: [colors]. No extra text or objects. Working note: Upload a straight-on product reference and a separate color reference when possible.
Localized event poster
Use when the wording must appear inside the image in a specific language.
Design a [ratio] poster for [event]. Display the exact headline '[headline]' and supporting line '[line]'. Use [visual style], high text contrast, and a simple reading order. Keep all other marks abstract and do not invent dates or sponsors. Working note: Generate one language at a time and proofread every character before use.
Consistent character scene
Use for a new scene while preserving a referenced person's main visual traits.
Using reference 1 as the character identity, show [character name] [action] in [setting]. Preserve [hair, clothing, and defining features]. Camera: [shot and angle]. Mood: [mood]. Do not add duplicate people or change the outfit colors. Working note: Reuse the same character name and reference order across the series.
Simple process infographic
Use for an explanatory visual with a small number of clearly ordered steps.
Create a [ratio] infographic explaining [process] in [number] steps. Use short labels, clear arrows, and one distinct visual per step. Exact title: '[title]'. Palette: [colors]. Keep the background plain and avoid decorative text. Working note: Verify facts separately; the model should arrange approved content, not be the source of the facts.
How to use Nano Banana 2 on Lovart ME
The same four-step process works for a first draft and for a reference-based edit.
- 01
Choose the input mode
Use Text to image for a new concept. Switch to Image to image when an existing subject, object, style, or composition must guide the result.
- 02
Write the brief and add references
Name the deliverable, subject, setting, composition, and constraints. For multiple references, explain the role of each image instead of asking the model to guess.
- 03
Set ratio, resolution, and format
Match the aspect ratio to the publishing channel. Start at 1K for exploration, then choose 2K or 4K for the approved direction. Select JPG or PNG.
- 04
Generate, inspect, and refine
Check subject identity, hands, small objects, typography, and edge details. Change one or two instructions at a time so you can tell which revision improved the image.
Nano Banana 2 questions
Answers reflect Google’s published model information and the options currently available in this Lovart ME workflow.
What is Nano Banana 2?
Nano Banana 2 is the product name for Google’s Gemini 3.1 Flash Image model. It is designed for fast image generation and editing, with support for detailed instructions, image references, text rendering, and multiple output formats.
Can I edit an existing photo?
Yes. Choose Image to image, upload one or more references, and describe what must change and what must stay fixed. Results still need review, especially for identity, text, and small details.
How many reference images can I upload?
The current Nano Banana 2 workflow supports up to 14 JPEG, PNG, or WebP reference images. Use only references that have a clear purpose, and describe that purpose in the prompt.
Does Nano Banana 2 support 4K images?
Yes. Lovart ME offers 1K, 2K, and 4K output for this model. Higher resolutions use more credits, so 1K is usually better for testing and 4K for an approved final direction.
Which aspect ratios are available?
You can use Auto plus 1:1, 2:3, 3:2, 1:4, 4:1, 3:4, 4:3, 4:5, 5:4, 1:8, 8:1, 9:16, 16:9, and 21:9.
Can it generate accurate text inside an image?
Google highlights improved text rendering and in-image translation. For production work, provide the exact copy in quotation marks, keep the layout simple, and proofread every word before publishing.
Is Nano Banana 2 free on Lovart ME?
New accounts may use included registration credits to try generation. Each job consumes credits according to the selected resolution; the live controls show the current cost before you generate.
Can I use the output commercially?
Usage depends on the platform terms, your source materials, and local law. Make sure you have rights to uploaded references and review current Lovart ME terms before using an image in commercial work.
Continue the image workflow
Use a more specialized route when the task needs a different model profile, a product-focused setup, or examples from the community gallery.