Nano Banana 2 Lite.

Google's Gemini 3.1 Flash Lite Image: text to image, and edits from 1 to 3 images, at 1k only. nano-banana-2-lite implements image.generate and image.edit on the Windpaint API and dashboard. It's a closed model run through Google Cloud Vertex AI, on purchased credits, from $0.036 per image.

Status
Live
Model id
nano-banana-2-lite
Capabilities
image.generate, image.edit
Takes
Prompt, plus 1 to 3 images to edit
Returns
One image
Tiers
1k
Runs on
Google Cloud Vertex AI
From
$0.036 per image (0.72 credits, metered)
Credits
Purchased only

Text to image and editing, closed model.

The catalog describes it as “Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image). 1K only.” It runs through Google Cloud Vertex AI, under Google's model id gemini-3.1-flash-lite-image.

Google's own page for the model has its full specs: Nano Banana 2 Lite on Google Cloud.

What it costs.

Metered prices follow the job's actual usage, such as output size or clip length. The listed price is the estimate held when you submit, and the charge never goes over it.

Nano Banana 2 Lite prices. 1 credit = $0.05.
ModelCapabilityTierCreditsUSDBilling
nano-banana-2-lite
image.generate
1k
0.72 per image
$0.036
Metered
nano-banana-2-lite
image.edit
1k
0.72 per edited image
$0.036
Metered

Closed models run on purchased credits only. The free signup credits work on the open-source models Windpaint serves. If your purchased balance can't cover the hold, the submit fails with 402 billing.insufficient_credits and nothing runs.

Failed and refused jobs aren't charged. Pricing covers credits and top-ups.

Request fields.

Submit to POST /v1/generation/capabilities/image.generate or POST /v1/generation/capabilities/image.edit. Anything outside these values returns 422.

Fields nano-banana-2-lite accepts.
FieldValuesDefaultNotes
model
nano-banana-2-lite
none
Name it; another model may be the capability's default.
prompt
string
none
Required. On image.edit, what to change.
inputs.images
1 to 3
none
image.edit only. Source and reference images, in the order the prompt refers to them.
resolution
1k
1k
The image tier.
aspect_ratio
any of the ten
1:1
Optional on image.edit.
output_format
png, jpeg, webp
png
Re-encoded on the server.
output_quality
1 to 100
90
jpeg and webp only.
webhook_url
https URL
none
Signed POST when the job finishes.

Call it.

The submit returns 202 with a request_id and a status_url. Send Prefer: wait to hold the response for up to 60 seconds and get the finished job back instead.

image.generatebash
curl -X POST https://api.windpaint.ai/v1/generation/capabilities/image.generate \  -H "Authorization: Bearer $WINDPAINT_API_KEY" \  -H "Content-Type: application/json" \  -d '{    "model": "nano-banana-2-lite",    "prompt": "a ceramic teapot on a linen tablecloth, soft window light",    "resolution": "1k",    "aspect_ratio": "3:2"  }'
image.editbash
curl -X POST https://api.windpaint.ai/v1/generation/capabilities/image.edit \  -H "Authorization: Bearer $WINDPAINT_API_KEY" \  -H "Content-Type: application/json" \  -d '{    "model": "nano-banana-2-lite",    "prompt": "make the teapot deep blue, keep everything else",    "inputs": {"images": ["d77a4c1e-9b2f-4a3d-8e6f-7a8b9c0d1e2f"]},    "resolution": "1k"  }'

Send an Idempotency-Key header so a retried submit returns the first job instead of billing a second. Each output is an asset in your project; its id can go straight into another job's inputs.

Nano Banana 2 Lite limitations.

Current limits on Windpaint, from the API's catalog.

  • One image per request; num_outputs above 1 returns 422.
  • Edits take no mask.
  • 1k only. Any other resolution returns 422.
  • The seed field isn't passed to the model, so the same request can give a different image.

Where it runs.

nano-banana-2-lite is a closed model from Google.

Windpaint runs it through Google Cloud Vertex AI. Your prompt and inputs for the job are sent to Google to make the output, and the output comes back to storage in your project. Windpaint doesn't train on your prompts, inputs or outputs.

Outputs are stored as assets in your project for your organization's retention period, 7 days by default. See Trust for moderation, data use and retention.

Questions.

How much does nano-banana-2-lite cost?

0.72 credits ($0.036) for 1k and 1k edit. The price is metered: it follows the job's actual usage, the listed price is the estimate held when you submit, and the charge never goes over it. One credit costs $0.05.

Can I use my free signup credits on it?

No. Nano Banana 2 Lite is a closed model, so it runs on purchased credits only. The 200 signup credits work on the open-source models Windpaint serves.

Where does Nano Banana 2 Lite run?

On Google Cloud Vertex AI, under Google's model id gemini-3.1-flash-lite-image. Your prompt and inputs for the job are sent to Google to make the output, which comes back to your project as an asset. Windpaint doesn't train on your prompts, inputs or outputs.

Am I charged if a job fails?

No. The price is held when you submit and released in full if the job fails or the model refuses the request.

Can Nano Banana 2 Lite edit an image?

Yes, on image.edit. Send 1 to 3 images in inputs.images and say what to change in the prompt. It takes no mask.

Is Nano Banana 2 Lite open source?

No. It's a closed model from Google, and Windpaint runs it through Google's own API. The open-source models Windpaint serves are z-image-turbo and flux-2-klein-4b for images, plus four audio models.

Start yours.

Sign up and get 200 free credits to spend in the first 90 days.