WP Nano Banana: AI image generation and editing that feels native to WordPress

Creating or editing images for a WordPress site still often means leaving the editor, firing up an external tool, exporting, then re-uploading and resizing — a workflow full of friction. WP Nano Banana aims to remove that waste: it brings AI image generation and editing into the exact places you already work inside WordPress (Media Library, media modal, post editor, page builders and any plugin that uses the core media popup).

You pick the model and provide the API key (BYOK), then generate or edit images using text prompts, uploaded reference files, or images from your Media Library — alone or combined. Results are saved straight to your library, so you never have to shuttle files back and forth.

See the project on GitHub: https://github.com/WP-Autoplugin/wp-banana


The Problem

Creating images for a site still feels like shoehorning a design workflow into a CMS. You open a separate app or web service, fiddle until the image looks right, export, upload, crop, and repeat. The back-and-forth adds time and context switches, increases the chance of mismatched sizes or color profiles, and pushes work outside the place where content is authored and published. Many plugins try to help — but they ship bespoke UIs that don’t feel like WordPress, hide features behind paywalls, or force you to use a hosted service (which may be fine, but isn’t always what teams want).

WP Nano Banana reduces that friction by putting generation and editing controls in the exact places you already reach for images: the Media Library, the core media modal (the dialog every plugin and editor uses), the post editor, and any plugin that relies on the WordPress attachment flow. No separate dashboards, no foreign design language — just tools that behave like native media actions.


What WP Nano Banana does (in one line)

It lets you generate and edit images with AI from inside WordPress, using text prompts, uploaded reference files, or images already in your Media Library, and saves results straight to the library.


Native-like integration

Designers and editors don’t want another panel to learn. They want a predictable flow. WP Nano Banana uses existing WordPress touchpoints instead of inventing new ones. That means the same media modal you use to pick a featured image will also let you generate or edit an image; the Media Library attachment view contains “Edit with AI”; and page builders that rely on the core media picker will show the same actions without special adapters.

The practical benefits are simple: less training, fewer surprises when updating editors or themes, and a consistent experience across plugins. It also makes adoption straightforward for teams — the plugin augments established flows rather than replacing them.


Generate and edit from basically anywhere in WP-Admin

WP Nano Banana surfaces generation and editing where images are already chosen or edited. Open the Media Library, click an attachment and choose Edit Image. In the post editor, the media popup offers a Generate Image button next to upload options. ACF/Metabox fields, WooCommerce product images, and other plugins that use the core media modal will automatically expose the same controls. There’s also a dedicated Media → Generate Image page for bulk ideation and drag-and-drop reference workflows.

That breadth matters because it eliminates the “export/import” loop. You can iterate a hero image from the post editor, produce product mockups from the product screen, or create a quick OG image without leaving the draft you’re writing.


Models for every task and budget

Not every image task needs the same model. Sometimes you want a fast, cheap draft to test composition and aspect ratio; other times you need a powerful editor for precise inpainting or photoreal fixes. WP Nano Banana lets you choose models that match the job. Examples include OpenAI’s gpt-image-1 for general generation and editing, Google Gemini variants (e.g. gemini-2.5-flash-image, a.k.a. Nano Banana) and a range of Replicate-hosted models (Seedream, Recraft, Ideogram, etc.) for editing and combining multiple images.

Because the plugin is BYOK (“bring your own key”), teams control cost and compliance: connect the provider you prefer, pick a cheaper model for batch work, and switch to a higher-quality model for final assets. The settings page shows available models and the currently configured provider so you can balance speed, quality, and price.


Create and edit with prompts + images (any combo)

WP Nano Banana supports mixed inputs. Write a text prompt, attach one or more reference images (from your computer or the Media Library), or do both. Want to turn a product shot into a lifestyle scene? Upload a background photo and attach the catalog image as a reference; add a prompt that instructs style, lighting, and crop. Need to remove a background or replace an object? Use targeted instructions and the model’s editing capabilities to get precise results.

The plugin keeps originals by default (or lets you save a copy), so you can safely iterate without losing source assets. Undo/redo and regenerate options make exploration quick: try variations, save the best, delete the rest.


Performance, privacy and developer notes

Performance: WP Nano Banana keeps a small footprint. The plugin sends the provided images and prompts to the configured API platform only when you trigger generation or editing. Asset handling uses WordPress’ native upload and attachment APIs so existing caching and CDN setups continue to work.

Privacy: you supply your API key, and the data only goes to the provider you choose. The plugin documents what’s sent (prompts, optional reference images) and when calls are made.

Developer-friendly: the codebase exposes hooks and filters for every major action, so you can adapt workflows, add server-side validation, or wire in logging. The plugin aims to be auditable and easy to extend.

For more details on provider policies and data handling, consult the provider docs you pick (for example: OpenAI Terms & Privacy, Google Terms & Privacy, Replicate Terms).


Real workflows — three quick examples

Blogger. While drafting, create an OG image without leaving the editor. Try several prompts, pick one, and set it as the featured image — all from the same draft.

E-commerce manager. Use a catalog product shot as a reference and generate lifestyle composites for marketing banners. Batch generate variations at cheaper models, then refine the best with a more capable editor.

Agency designer. Maintain brand consistency by storing reference images in the library and feeding them to the generator. Iterations stay versioned in the Media Library and are easy to swap into client pages.


What WP Nano Banana isn’t

It’s not a hosted managed service that stores your assets or locks features behind tiers. It doesn’t try to replace full desktop editors for complex work — professional tools still have their place. Instead, it’s an in-place tool for ideation, quick fixes, and integration with content workflows. You’ll need API keys for the providers you want to use and to be mindful of provider limits and pricing.


Roadmap & where to try it

Planned improvements include inpaint/outpaint/masking tools, batch edit flows, more model integrations, and optional usage analytics. You can grab the plugin, file issues, or contribute on GitHub: https://github.com/WP-Autoplugin/wp-banana


WP Nano Banana – Frequently Asked Questions

Setup & API Keys

Where do I add my API keys?
Go to Settings → WP Nano Banana and paste your provider key(s). You can also load keys from wp-config.php – see the Readme file for details.
Which providers are supported?
OpenAI, Google (Gemini), and Replicate. You bring your own key for whichever provider(s) you prefer.
Do I need a paid account or credits?
Yes. All listed providers charge for image generation/editing. If you see HTTP 429 (“rate limit”/“quota exceeded”) with a “free” key, it usually means billing isn’t enabled or you’ve exhausted free/intro credits. Add a payment method or credits on your provider dashboard, then retry.

Where to Find the Features (UI)

How do I generate a new image?
Use any of the following:
  • Media → Generate Image: a dedicated page with Prompt, Provider, Model, Aspect Ratio, and an attach image paperclip button. Supports drag-and-drop.
  • Media → Library: click the Generate Image action button next to “Add Media File” to expand the Generate panel.
  • Media modal (the core popup): a new Generate Image tab appears alongside “Upload files” and “Media Library”.
How do I edit an existing image with AI?
In the media modal or Media Library, click an image to open Attachment detailsEdit Image (WordPress editor). There you’ll see an AI Edit button that expands the Edit panel. WordPress’ default Undo/Redo buttons work for AI changes, too.

Models & Providers

Which models are available for generation?

Gemini

  • gemini-2.5-flash-image-preview

OpenAI

  • gpt-image-1

Replicate

  • google/gemini-2.5-flash-image
  • google/imagen-4
  • google/imagen-4-ultra
  • google/imagen-4-fast
  • black-forest-labs/flux-1.1-pro
  • black-forest-labs/flux-dev
  • black-forest-labs/flux-schnell
  • recraft-ai/recraft-v3
  • ideogram-ai/ideogram-v3-turbo
  • ideogram-ai/ideogram-v3-quality
  • ideogram-ai/ideogram-v3-balanced
  • bytedance/seedream-4
  • qwen/qwen-image
Which models are available for editing?

Gemini

  • gemini-2.5-flash-image-preview

OpenAI

  • gpt-image-1

Replicate

  • qwen/qwen-image-edit
  • bytedance/seededit-3.0
  • bytedance/seedream-4
  • google/nano-banana
  • black-forest-labs/flux-kontext-max
  • black-forest-labs/flux-kontext-dev
How should I choose a model?
For quick drafts or ideation, pick faster/cheaper models. For precise edits (subject replacement, detailed retouch), choose models known for editing quality. If a model rejects content or returns low detail, try another from the same provider or switch to OpenAI/Gemini/Replicate equivalents.
Note that editing/combining multiple reference images is not supported by all models.

Using Prompts & Reference Images

Can I attach reference images?
Yes. On the Generate page, use the paperclip button or drag & drop. In the media modal, use the Generate Image tab’s attach control. For editing, open AI Edit from the WordPress image editor and describe the change you want (you can also attach references where supported by the model).
Why do results vary between runs?
Image models are stochastic. Small prompt changes (or none at all) can still yield different outputs. If a provider supports seeds or strength parameters, try those to improve consistency.

Errors & Troubleshooting

I’m getting HTTP 429. What does it mean and how do I fix it?
429 usually means rate limit or quota/billing issues. Common causes:
  • Free key without billing/credits (very common) → enable billing/add credits.
  • Too many requests in a short window → wait and retry; avoid rapid multi-generation.
  • High concurrency from the same server/IP → stagger requests.
After fixing billing or waiting out the window, try a single image first to confirm.
I’m getting HTTP 400 (Bad Request).
The provider rejected your parameters. Check:
  • Model ID is valid for your chosen provider and task (generate vs edit).
  • Prompt is present and within length limits.
  • Aspect ratio is supported by that model (for Generate only).
  • Reference images meet size/type limits; try smaller RGB PNG/JPG.
  • Prompt doesn’t violate provider safety rules (rephrase if needed).

Output & Files

Where are generated/edited images saved?
In your standard WordPress uploads directory (wp-content/uploads/YYYY/MM) as normal attachments. They behave like any other Media Library item.
Where are temporary states of the images saved for the undo/redo functionality?
They are stored in the database as transients. They are automatically cleared after a while. You can also manually manage transients with plugins.
Can I control aspect ratio?
Yes, on the Generate panel (not on AI Edit). Note that some models only accept certain sizes/ratios; invalid combinations can cause 400/422 errors.
Can I control output format?
Yes, in the Plugin Settings under Defaults. Supported formats are PNG, JPG, and WEBP. Note that some models only support specific formats; unsupported choices may lead to errors.

Privacy & Compliance

What data is sent to providers?
Only what’s needed for the request: your prompt and any attached reference images. Requests are sent only when you trigger Generate or AI Edit. Review your provider’s terms to understand retention/training settings and update your site’s privacy policy accordingly.

Compatibility & Requirements

Which parts of WordPress does it work with?
Anywhere the core media modal is used: Gutenberg, Classic Editor, many page builders, WooCommerce product images, ACF/Meta Box fields, and more.
Minimum requirements?
WordPress 6.0+, PHP 7.4+, Imagick extension enabled, and an API key for a supported provider (OpenAI, Google, or Replicate).