Internal + Public

Getting started with Image Studio

How Image Studio is laid out, how to make your first image, which of the 24 models to pick, what every setting does, and how the Edit menu sends a result to Image Editing, Video Studio, Filter Gallery or Lip Sync.

Updated 2026-10-31 Image Studio imagesgetting-startedtutorialmodelsreference-imagesedit-menuanimate

The Image Studio layout

Open Image Studio. Three panels, left to right:

  1. Left: what to make. The Prompt box, the Image Style tiles, the AI Enhance Prompt checkbox and the AI Model card. Click Change on the card to open the model picker.
  2. Middle: Settings. Only the controls the chosen model has: aspect ratio or image size, resolution, quality, number of images, reference images, output format. Underneath are the Estimated Cost box and the Generate Images button, which always shows the live price.
  3. Right: Results. Your images in their true shape, with their actions.

Your first image

  1. Leave the model on Grok Imagine (the default: fast, 2 images per run, 6 credits).
  2. Type a prompt. Lead with the subject, then the setting, the light and the mood.
  3. Pick an aspect ratio: 1:1 for a cover or post, 9:16 for a story, 16:9 for a banner.
  4. Click Generate Images. Most models answer in 5 to 30 seconds.
  5. The images appear on the right and are already saved to your Gallery. There is nothing to save by hand.

If the first result is not right, run it again. Every run is different, and the second or third is often the keeper.

Writing the prompt

Structure: [subject], [setting], [style], [light], [mood], [technical details]

"A woman with neon pink hair in a rainy Tokyo alley, cinematic photography, soft backlight from a vending machine, melancholy, shallow depth of field"

Image Style tiles: None, Photorealistic, Abstract, Cyberpunk, Retro, Minimal, Surreal, Cinematic, Artistic, Anime, Watercolor, Oil Painting, Digital Art, 3D Render. A tile adds that style's keywords to the prompt. Use None when the prompt already describes the look.

AI Enhance Prompt: asks the model to expand a short prompt with more detail. Only models that support prompt expansion use it; the rest ignore the box.

Choosing a model

The picker has tabs All, Favorites, Budget, Standard and Premium. Click the heart on a tile to pin it; favorites follow the account to every device. Badges: NEW (recent), PRO (higher-end), SPICY (looser content filters; images are tagged as adult content).

You want Start with
Many ideas, cheaply Flux Schnell (1 cr), Z-Image Turbo (2 cr)
Everyday all-rounder Grok Imagine, Nano Banana 2 Lite (6 cr)
Exact text on the image GPT Image 2, GPT Image 2.5 Flare or Sunburst
Photoreal people and products Flux 2 Pro
A character or product placed in a new scene GPT Image 2.5, Nano Banana 2, Nano Banana 2 HD (14 to 16 references)
Print and large format Nano Banana 2 HD, Nano Banana 2, Seedream V4 (true 4096 px)
Several variations in one run Wan 2.7 Image (up to 4), Seedream V4 (up to 6)

All 24 models (starting prices checked 2026-10-07; the Generate button always shows the current price):

Tier Model From Reference images Settings
Budget Flux Schnell 1 cr none Image size
Budget Z-Image Turbo 2 cr none Image size
Budget Qwen Image 2-6 cr 1 Image size
Standard GPT Image 1.5 6 cr 16 Aspect ratio
Standard Grok Imagine 6-7 cr 5 Aspect ratio, Quality Standard or Pro, 2 images
Standard Grok Imagine 2.0 6 cr 3 Aspect ratio
Standard Nano Banana 6 cr none Aspect ratio
Standard Nano Banana 2 Lite 6 cr 10 Aspect ratio incl. Auto
Standard Flux 2 Pro 7-10 cr 8 Aspect ratio incl. Auto, 1K or 2K
Standard Flux Kontext Pro 7 cr 1 Aspect ratio
Standard Seedream V4 7 cr per image 10 Image size, 1K/2K/4K, up to 6 images
Standard Nano Banana 2 8-18 cr 14 Aspect ratio incl. Auto, 1K/2K/4K, PNG or JPG
Standard GPT Image 2 9-12 cr 16 Aspect ratio incl. Auto, 1K or 2K
Standard GPT Image 2.5 Flare 9-23 cr 16 Aspect ratio incl. Auto, 1K/2K/4K
Standard GPT Image 2.5 Sunburst 9-23 cr 16 Aspect ratio incl. Auto, 1K/2K/4K
Standard Qwen Image 3.0 9 cr none Image size
Standard Qwen Image Plus 9 cr none Image size
Standard Wan 2.5 (SPICY) 9 cr 1 Image size
Standard Nano Banana 2 HD 12-23 cr 14 Aspect ratio incl. Auto, 1K/2K/4K, PNG or JPG
Standard Flux Kontext Max 14 cr 1 Aspect ratio
Standard Grok Imagine Quality 14 cr 8 Aspect ratio incl. Auto, 1K or 2K
Standard Wan 2.7 Image 7 cr per image 10 Aspect ratio, 1K or 2K, up to 4 images
Premium Flux 2 Flex 20-34 cr 8 Aspect ratio incl. Auto, 1K or 2K
Premium Nano Banana Pro 26-34 cr 14 Aspect ratio, 1K/2K/4K, PNG or JPG

Not every model is from the same supplier. If a model is unavailable, another maker in the same tier is usually still running.

What each setting does

Using reference images

Add references with Upload, From Gallery or UGC character, or drop files on the slots. The counter shows the model's limit (for example 0 of 16). Single-reference models replace the reference when you add another; models with no slots take none.

After a run, the results card shows Built from with the references that shaped that batch.

After you generate

Each batch is one results card:

On each image: Save (shows it is saved), Download, Share to Feed, Share (to socials), Edit (cyan wand) and Open large.

The Next step strip

Under a fresh batch, a Next step strip offers one-tap tiles, each with image 1 of the batch already loaded:

Close the strip with the x if you do not need it.

The Edit menu: send a result to another app

The cyan Edit wand on any result opens a menu. Each item opens the destination with the image already loaded:

Item Destination
Upscale Image Editing, Upscale selected
Remove background Image Editing, Remove BG selected
AI Edit Image Editing, AI Edit selected
Reframe Image Editing, Reframe selected
All edit tools Image Editing
Animate in Video Studio Video Studio with the image as the first frame; the model switches to its image-to-video version
Filter Gallery Filter Gallery
Make it talk (Lip Sync) Lip Sync with the image as the face

The image is already saved, so leaving Image Studio loses nothing. Image Editing's tools are explained in Image Editing.

Common questions and fixes

When generations fail

See generation-failures. If the credits came back, it worked as designed. If they did not, report the generation ID.