Image 2 Updates

This update brings a model upgrade, a price cut and 4K for the top quality levels.

What's new

• Nano Banana 2 is now Nano Banana 2.1, still at 4 / 5 / 6 credits. Older images in your history can still be viewed and reused.
• Nano Banana Pro now costs 5 / 5 / 7 credits for 1K / 2K / 4K, down from 7 / 7 / 9.
• Ultra and Extreme quality on GPT Image 2.5 now support 4K, at 52 and 115 credits. 4K at these levels takes longer to generate.
• Fixed: after you change settings, the credit hint under the generate button switches back to the estimate for your next image.

Halloween is almost here, and this update adds a page made for it.

What's new

• The new AI Halloween Costume Generator turns a selfie into a Halloween photo: upload a photo, pick a costume, and generate.
• The page has 8 Halloween costume presets. Clicking a card fills in the prompt in the generator on the same page and keeps the photo you've already uploaded.
• GPT Image 2.5 Flare is the default model, and you can switch to others. A "Which model should I pick?" section explains the options.

This update makes viewing, downloading and topping up smoother.

What's new

• Results show a compressed preview first so they load faster. If a preview fails to load, you can reload it.
• A new Download original button sits in the result area, so it's easy to find on phones and tablets too.
• When your balance is too low, the generate button changes to "Not enough credits, top up" and shows how many credits you need.
• Your prompt, aspect ratio, resolution, model, reference images and other settings are saved before you go to top up, and restored when you come back.

This update lets you trade speed and credits for detail.

What's new

• The generator has a new Quality option: Standard, Enhanced, Fine, Ultra and Extreme. GPT Image 2.5 Flare and Sunburst support all five, and GPT Image 2 offers Standard, Fine and Extreme.
• Credits per image (1K / 2K / 4K): Standard 2 / 4 / 6 (unchanged), Enhanced 3 / 6 / 8, Fine 10 / 18 / 30, Ultra 16 / 32, Extreme 35 / 70. Each level costs the same on every model.
• Levels above Standard are for accounts that have bought credits or a subscription. On an unpaid account you'll be asked to top up first, and no credits are used.
• Image-to-image is billed by actual complexity, so levels above Standard need a balance of at least twice the listed price. The quality hint shows the balance you need.
• Your quality choice is saved with drafts and history, and comes back when you reuse settings from history.

This update adds two Google image models to the generator.

What's new

• The model menu now includes Nano Banana 2 (Gemini 3.1 Flash Image) and Nano Banana Pro (Gemini 3 Pro Image) for text-to-image and image-to-image.
• Image-to-image accepts up to 14 reference images, with 10 aspect ratios and 1K, 2K or 4K output.
• Fixed credits: Nano Banana 2 uses 4 / 5 / 6 credits and Nano Banana Pro 7 / 7 / 9, the same for text-to-image and image-to-image. Failed generations are not charged.
• A new Nano Banana 2 page explains the difference between the two models and how to write prompts for them.
• History now shows the name of the model that created each image.

This update focuses on speed and reliability.

What's new

• History thumbnails now load as small images, so the list opens much faster and uses less data. Details and enlarged views still use the original image.
• Your credit balance now shows a loading state instead of briefly showing zero credits and a purchase prompt.

Click any image in My History to see its full generation details and pick up where you left off.

What's new

• A new History details view shows the result, the original reference images and the generation settings.
• Click Reuse original settings to restore the prompt, reference images and all parameters, then edit before generating.
• Quick actions in the list let you reuse settings, continue creating or download in one click.
• Previewing never changes your current draft, and you can undo after reusing settings.

Refining or extending earlier work no longer means re-uploading references and rebuilding your settings.

Some reference images cannot be read because of their format or encoding, which causes the generation to fail. You no longer need to convert and re-upload them yourself.

What's new

• When a reference image cannot be read, click Optimize and retry to process it and generate again.
• Your original upload stays unchanged; the optimized image may show minor color or detail differences.

If optimization fails, no new generation is submitted and no credits are charged.

You can now use GPT Image 2.5 on Image2 to create images from text or reference images.

What's new

• A new GPT Image 2.5 page, with Flare selected by default.
• Switch to Sunburst from the generator's model menu, or go back to GPT Image 2 at any time.
• Choose the aspect ratio, 1K / 2K / 4K resolution, and PNG, JPEG or WebP output.
• Credit rates are the same as GPT Image 2, and failed generations are not charged.
• History keeps the exact model used for each image, so you can review and compare results later.

To compare Flare and Sunburst, keep the prompt, reference images and output settings the same, then choose the model that works best for your task.

We increased the prompt length limit and made it easier to understand why generation cannot start or why a task has failed.

What's new

• The prompt limit has increased from 2,000 to 32,000 weighted characters, with Han characters counting as two.
• The prompt field now shows the current character count and maximum length in real time.
• When generation cannot start, the interface explains whether the prompt is empty or too long, a reference image is still uploading, or an upload has failed.
• Long error details are now scrollable, so the full message remains available instead of being cut off after a few lines.

Whether you are describing a complex scene or troubleshooting an upload or generation issue, the interface now provides more immediate and useful feedback.

You can now generate transparent-background images directly with GPT-Image-2, without removing the background afterward.

What's new

• Choose from Auto, Opaque, and Transparent background modes.
• Transparent mode generates PNG or WebP images with an alpha channel.
• If JPEG is selected, choosing a transparent background automatically switches the output format to PNG.
• Create logos, product assets, stickers, icons, and other design elements that are ready to place in any layout.

Transparency is submitted as a generation setting rather than relying on prompt wording alone, making the output more consistent and controllable.

We updated GPT-Image-2 pricing to make credit usage clearer and better aligned with each generation.

What changed

• Image-to-image generation now starts at 2 credits for 1K, 4 credits for 2K, and 6 credits for 4K.
• The final image-to-image cost is confirmed after generation based on the reference images, image size, and task complexity.
• The 4K text-to-image price has been reduced from 8 credits to 6 credits.
• Failed generations are not charged.
• The generator now shows the starting price, billing confirmation, and final credits used.

You can review the latest rates in the Credit Usage section.

You can now create an account and sign in with a verification code sent directly to your email.

What is new

• Request a one-time verification code from the sign-up or sign-in page.
• Enter the code on the verification page without opening a magic link.
• Receive localized email content based on your selected language.
• Get clearer feedback when requesting, entering, or resending a code.
• Benefit from improved request limits and security protections.

Adding reference images is now faster: copy an image from another app or webpage, then paste it directly into the uploader.

What is new

• Use the new Paste button to insert an image from your clipboard.
• Press Ctrl+V on Windows or Command+V on macOS.
• See clearer upload progress and error messages.
• Clear all reference images in one action.
• Get helpful feedback when the clipboard is empty, access is denied, or a file is unsupported.

The existing file picker and drag-and-drop upload options are still available.

Your previous generations are now easier to find and reuse.

What changed

• Open My History directly from the account menu in the top-right corner.
• View the images you generated in one place.
• Select a previous result to reuse its prompt and generation settings.
• See clearer guidance explaining where generated images are saved.

This makes it easier to continue an earlier idea or create a new variation without rebuilding your settings.