On this page
01
Use cases
What this model is useful for
Create a new visual
Describe the subject, setting, light, format and space needed for layout. Do not rely on the model for important small text; add it later in a design editor.
Replace a product background
Upload the source photo and list what must stay unchanged: shape, color, artwork, viewpoint and object count. Compare source and result side by side.
Build an illustration series
Repeat palette, lighting, viewpoint and size in every prompt. One successful image does not guarantee character or object consistency across a series.
02
Examples
Prompts you can try
A sample brief — add your own inputs. This page does not generate content or charge your balance.
Prompt for a future run — no source image, result or actual charge has been preserved yet.
Replace a mug background
A reproducible example needs an authorized mug photo. Preserve it with the result and disclose any manual retouching.
Use the uploaded mug image as the source. Replace only the background with plain light gray. Preserve the handle shape, glaze color, existing artwork and mug position. Add a soft contact shadow. Do not add a second mug, text or watermarks. Ask for the image if none is attached.Prompt for a future run — no completed image or verified settings are attached yet.
A new lamp illustration
Choose an available square format before running. Review composition, empty headline space and accidental text.
Create a square editorial illustration: a small yellow desk lamp on a blue cabinet in an evening room, with soft warm light and a simple modern style. Leave clear space at the top. No text, logos or people.Make the prompt your own
- For editing, upload your image and replace the scene description with the changes you want.
- If a particular object matters, state which details must not change.
How to check the result
- In the edit, the background changes while the handle, glaze, artwork and mug position remain intact.
- The product outline is not cropped, and the contact shadow matches the surface.
- The new illustration is square, leaves space at the top and contains no text, logos or people.
- A published result preserves the source, exact prompt, settings, file, model ID, date, cost and post-processing.
03
Pricing / USD
CLODEx public pricing
Use the CLODEx price shown for the selected mode. A retry or another variation creates additional usage; Google API pricing does not replace the CLODEx rate.
Full pricing details04
Get started
How to get started
Check the ID and available interface
Confirm that nano-banana-2 is available to your key or appears in the CLODEx image interface. Do not replace the ID with a Google SDK model name.
Choose generation or editing
A new image can start from text. Background replacement needs an owned or authorized source uploaded through a supported CLODEx route.
Review before publishing
Inspect the full-resolution image, compare the product with the source and check text, logos, fingers, edges and fine details.
Check model access
curl --fail-with-body 'https://clodex.xyz/v1/models' \
-H "Authorization: Bearer $CLODEX_API_KEY"This example only reads the catalog and does not generate content. Choose job settings in Studio; the exact cost is shown before submission.
How to get an API key05
Specifications
Capabilities and limitations
- Model ID
nano-banana-2- API protocols
- OpenAI-compatible API
Only parameters declared by the integration catalog are shown. An omitted parameter is not a promise of support.
Model and integration
How the names relate
Google calls Gemini 3.1 Flash Image Nano Banana 2. CLODEx exposes the separate public ID nano-banana-2, which should be used as written.
Two different workflows
Text generation creates a new scene. Editing must preserve a specific source object, so it requires a reference and stricter review.
What to consider
The CLODEx route differs from the Google SDK
The upstream model name does not imply the same request body, upload flow or parameters. Use only the CLODEx format.
Exact product preservation is not guaranteed
Handles, artwork, proportions or edges can change. Reject catalog images that alter the product itself.
Text and logos need review
Small lettering may be distorted and stray glyphs may appear. Add important copy in a conventional editor.
Before you submit
The page shows prompts, not completed results. Editing requires a source file and a supported upload path; without the image, the model cannot preserve a specific mug's shape.
06
Related models
Models to compare
GPT Image 2
gpt-image-2Use the same source and brief. Compare product geometry, edges, accidental text, available modes and actual cost per accepted variant.
07
FAQ
Model FAQ
Can it create an image from text alone?
Yes, when the selected CLODEx route offers generation. Text alone cannot preserve a specific product.
Can it replace a photo background?
Only through a supported source-image workflow. List every detail that must remain unchanged and compare the result with the original.
Why are there no completed images here?
No reproducible runs with sources, settings and cost have been preserved yet. The page shows honest prompts instead of fabricated results.
Is generated text reliable?
Do not rely on error-free typography. Add headlines, prices and required copy after generation.
How is cost calculated?
By the public CLODEx rate for the selected mode. Each additional variation or retry is billed separately.
08
