How to Use Nano Banana 2 in ChatGPT

You can call Nano Banana 2 directly inside ChatGPT through APIPod’s remote MCP. After you connect the MCP server, name the model nano-banana-2 in your prompt and describe the scene. APIPod runs the Nano Banana 2 API and returns the image URL into the chat. You do not need a separate Google Gemini account, and you do not paste an API key into ChatGPT.

Nano Banana 2 fits product shots, marketing assets, greeting cards, banners, and fast iteration when you need subject consistency across reference images. This path supports 512px / 1K / 2K / 4K, 14 aspect ratios, and up to 14 reference images. Full parameters: model page and docs.

Connect Nano Banana 2 in ChatGPT#

1. Open ChatGPT on the web → Settings → Plugins → Browser Directory (open the plugins page).

2. On the plugins page, click Add in the top right → Add custom MCP Server to start adding a custom MCP server.

3. In the dialog, fill in the MCP basics. The most important field is the Connection URL:

   TextCopy

   https://api.apipod.ai/mcp

   Under Authentication, choose OAuth.

   Under Advanced OAuth settings, set OIDC scopes supported to profile and email. Everything else is filled automatically — leave it as is.

4. Accept the risk notice and click Create a plugin. A connector dialog opens; click Connect to start OAuth for the MCP server:

5. This step matters: you authorize on www.apipod.ai. Later generations bill against that account’s balance. Click Allow Access on the APIPod consent page.

6. After authorization, open the Personal tab on the plugins page to see the custom plugin you created. Click it to open the plugin detail page:

7. Click Try in chat in the top right to jump into a chat with the plugin enabled. In your prompt, say “use nano-banana-2”, give aspect ratio and quality, and describe the scene to call APIPod’s remote MCP and generate a nano-banana-2 image:

8. The result shows up in the chat. You can also find the matching job in the APIPod console task list:

Note that your ChatGPT plan and your APIPod balance are separate. APIPod charges in USD per successful image; provider-side failures refund the reservation automatically.

A product-ad prompt you can copy#

@APIPod Use nano-banana-2 for a 1:1 image at 2K.
Use this public HTTPS URL as a product reference (keep bottle shape, cap, and label):
https://example.com/your-perfume.jpg
Compose a new hero shot: bottle on wet black stone, warm side light, dark background, subtle mist, clear label typography.
Run estimate_cost first and wait for my OK before generate_image.

A 2K square take lists at $0.06 per request. Check tiny label text after generation; for campaign typography, refine in a second edit or finish titles in design software.

A useful prompt shape: subject → action / pose → setting → camera / framing → lighting → style → on-image text (if any). Change only one or two variables per take so you can tell what moved the result.

Text-only, edit, or multi-reference?#

Tell ChatGPT which mode you want. Nano Banana 2 accepts text plus optional image_urls (up to 14).

Over MCP, references must be public HTTPS URLs. The connector does not upload local files — upload in the APIPod console first, then paste the returned URL. The MCP path does not accept base64.

Nano Banana 2 pricing in ChatGPT#

APIPod list prices per successful request. Final charge follows settlement.

Aspect ratio does not change this model’s unit price. Every new generate_image is a new charge — have ChatGPT run estimate_cost first (and get_balance if you want).

Nearby catalog options like Seedream 5.0 Flash or GPT Image are cost comparisons only, not quality rankings. Live prices: Nano Banana 2.

Habits that save money and avoid mistakes#

• Model id: name nano-banana-2 explicitly so ChatGPT does not fall back to a generic image default.
• Prompt length: keep under about 3993 characters.
• Aspect ratios: 1:1, 1:4, 1:8, 2:3, 3:2, 3:4, 4:1, 4:3, 4:5, 5:4, 8:1, 9:16, 16:9, 21:9 (default 1:1).
• Quality: default 1K; move to 2K / 4K when you need print or crop headroom.
• Latency: images usually land in about 10–40 seconds; ChatGPT polls check_job until permanent URLs appear.
• Billing: charges hit the APIPod account you authorized with OAuth. There is no MCP subscription fee; rate limits match your REST plan.

Keep the next iteration specific#

Check identity, product shape, text, and crop first, then change one or two things: “Keep the bottle and lighting, tighten the label text, switch to 9:16.” Each new take is a new paid generation.

FAQ#

Can ChatGPT generate Nano Banana 2 images?#

Yes. Through APIPod remote MCP: add the custom MCP server https://api.apipod.ai/mcp, sign in with your APIPod account via OAuth, @ the connector, and name nano-banana-2.

How much does Nano Banana 2 cost in ChatGPT?#

APIPod list price: $0.04 for 512px / 1K, $0.06 for 2K, $0.08 for 4K, per successful request. Run estimate_cost with your actual quality before generating.

Do I need to paste an API key into ChatGPT?#

No. Auth is OAuth 2.1 with PKCE — nothing to paste in the client. The server bills the account you signed in with.

Can I edit an image with Nano Banana 2?#

Yes. Put a public HTTPS URL first in image_urls and describe the edit. For multi-turn person or product consistency, add more references (up to 14).

How do I use Nano Banana 2.1?#

Same ChatGPT connection. Name nano-banana-2.1 for the newer Flash Image tier (1K / 2K / 4K; that id has no 512px tier). Model page: nano-banana-2.1.