Image Studio

Image Studio (/create-image) is one page for everything you do with images: create a picture from a description, edit or combine photos, increase resolution, prepare product photos for marketplaces and build a carousel. It's made for content creators, social media managers, marketplace sellers and designers, and there's nothing to install.

Studio tabs

• Text to image

  Price: from 1 credit

  Models: Nano Banana 2, GPT Image 2, Z-Image and 19 more models

  A new image from your description.

• Edit photo

  Price: FLUX Kontext Pro — 3 credits

  Models: Nano Banana 2, FLUX Kontext Pro, GPT Image 2 and others

  Change an uploaded photo with a description: background, clothing, lighting, details.

• Combine photos

  Price: Seedream 4.5 — 3 credits

  Models: Nano Banana 2, GPT Image 2, Seedream 4.5 and others

  A new scene from 2–8 photos: people, objects, a background, a style reference.

• Enhance quality

  Price: 3 credits

  Models: Clipia Photo Enhance

  Increases photo resolution up to 4× with no settings to tweak.

• Marketplace photo

  Price: no charge on this tab

  A single product photo shoot: your product photos and wishes go to Marketplace Studio. The tab is shown only to accounts with access to Marketplace Studio.

• Carousel

  Price: content at no charge

  Models: Nano Banana 2 for illustrations

  A multi-slide carousel editor for social networks.

Each tab has its own address: Text to image opens at /create-image, the other tabs at /create-image/<tab>. You can share a link that opens the studio on the right tab — for example, /create-image/ai-photo-editor.

How to create an image

Each launch creates one image. To compare variations, press Try again or run the same prompt on another model. Every launch is paid separately.

Settings

When you switch models, the aspect ratio and quality stay if the new model supports them. The negative prompt and seed are cleared if the new model doesn't accept them.

How to choose a model

The model picker only lists models that work on the current tab. Recommended models come first:

Filters above the list narrow it down:

• Up to 4K — models that support 4K.
• Renders text — models that draw readable lettering: Nano Banana 2, Nano Banana Pro, GPT Image 2 and 2.5, Seedream 5.0 Pro, Qwen Image 2.
• Covered by welcome credits — models you can run with the 5 credits you get at sign-up.
• Fits your balance — models your current balance can pay for.
• Best quality — GPT Image 2.5 Sunburst, Nano Banana Pro, Seedream 5.0 Pro, FLUX 2 Pro, Nano Banana 2.
• New — recently added models.

You can sort the list: Recommended, Lowest price first, Highest price first. A model card shows a "from N credits" price; the exact price for your settings is on the launch button. For a detailed comparison, see Image models.

Which models work on which tab

Text to image

The main studio mode: you write a description and the model draws it. Start with an inexpensive model, then repeat a good idea at a higher resolution or on another model.

Model specifics:

• GPT Image 2: the Auto aspect ratio is only available at 1K, and 1:1 isn't available at 4K.
• Midjourney V7 works from text only, without uploaded photos.
• Imagen 4 Fast and Ultra accept a negative prompt.

Edit photo

Upload a photo and write what to change: background, clothing, lighting, details. The model tries to keep everything else.

1. Upload the main photo. It's required.
2. If the model accepts several photos, add extra ones: they can set the background, clothing or style.
3. In one or two sentences, describe what to change and what to keep.
4. Press Edit photo.

If the result isn't right, make the description more specific, for example "keep the face and clothing unchanged", and run it again.

Combine photos

Upload several shots — people, objects, a background, a style reference — and describe the final scene: who stands where, what they wear, what the background is. Press Combine photos.

• You need at least 2 photos. The maximum is 8, or less if the model's limit is lower. The counter under the photos shows how many more you can add.
• The number of photos doesn't change the price: you pay for one final image.
• If the model mixes up details, name them directly: "the jacket from the second photo".

Uploads and limits

Each file can be up to 10 MB. JPG, PNG, WebP, HEIC, AVIF, TIFF, BMP and GIF are accepted; less common formats are converted automatically. You need to sign in to upload photos.

The Combine photos tab has an overall cap of 8 photos, even if the model accepts more.

Enhance quality

This tab increases photo resolution up to 4×. Proportions are kept and there's nothing to configure. It runs on Clipia Photo Enhance at a fixed price of 3 credits per photo.

1. Upload one photo: JPG, PNG, WebP, HEIC, AVIF, TIFF, BMP or GIF up to 10 MB.
2. Press Enhance quality.
3. Download the result from the canvas or from My tasks.

You don't need to download and re-upload an image made in the studio: choose Use in… → Enhance quality. If the page was reloaded, upload the photo for enhancement again.

Enhancement makes a photo sharper, but it won't invent what isn't there. To choose 2× or 4×, use the Topaz mode or enhance a video, go to Enhance and upscale.

Marketplace photo

This tab prepares a single product photo shoot — product shots without text — and hands it to Marketplace Studio. It is not a card set: cards with text for Wildberries and Ozon are made in Marketplace Studio itself. The tab is shown only to accounts that have access to Marketplace Studio.

Upload up to 6 product photos (JPEG, PNG or WebP, up to 10 MB each); the first becomes the main one. Describe your wishes in 3 to 1,000 characters — where to shoot the product, what lighting and mood. Open marketplace studio sends the photos and text to Marketplace Studio, where you pick scenes, see the price of the set and start the shoot.

No credits are charged on this tab. Sharp product photos on a plain background work best. Learn more in Marketplace cards.

Carousel

This tab opens the carousel editor: a slide plan and copy from your topic, styles, layouts and export. Everything about it is in Carousel Studio.

While it's generating

After you press the button, the task appears in My tasks as Queued, then Creating. The canvas shows "Creating · usually under two minutes" and a progress indicator.

• You can close the page: the image will stay in My tasks and My works.
• Pressing the button again while a launch is in progress doesn't create a second task.
• If there's no result after 15 minutes, the canvas stops waiting and suggests following the task in My tasks. The generation itself isn't cancelled.
• My tasks shows up to 20 recent launches. You can filter them by Video, Photo and Audio.

The text and settings of each tab are saved in your browser as a draft, separately for each account. Uploaded photos aren't kept in the draft.

What to do with the result

• Download — save the file to your device.
• Use in… — pass the image to Edit photo, Combine photos or Enhance quality without uploading it again.
• Try again — run the same request once more for another variation. Credits are charged again.

Every finished image is saved in My works.

Pricing

The price depends on the model and the chosen resolution or quality. The number of uploaded photos doesn't affect it.

You get 5 welcome credits when you sign up — enough for one Nano Banana 2 image. Credits and plans are explained in Credits and plans.

Troubleshooting

Tips

• The more specific the description, the more accurate the result: name the subject, place, lighting and mood.
• Need text in the image? Pick a model tagged Renders text and put the text in quotes.
• Explore ideas on an inexpensive model such as Z-Image or Grok Imagine, and make the final version on a stronger model at a higher resolution.
• When editing, the Auto or Match photo aspect ratio keeps the source proportions.
• Use a sharp photo for edits: the crisper the shot, the cleaner the result.
• Only combine photos of people with their consent.
• To get close variations on Imagen 4, Qwen Image 2 or Wan 2.7 Image, set the seed yourself and keep it fixed.