AI Image Playground
The AI Image Playground is a dedicated workspace for turning a prompt into study visuals, diagrams, and illustrations. Unlike images you generate inside a chat, the Playground is a focused canvas for image work — pick a quality level, an aspect ratio, choose one or more image models, and generate side-by-side variations from the same prompt.
The Playground is in beta, so each plan has a small daily image cap while we tune quality and speed.
Opening the Playground
Open the AI Image Playground tile from your home create menu — it shows a small New badge so it's easy to spot. The Playground opens with a prompt box at the bottom of the screen and a clean canvas above where your generated images will appear.
If you haven't typed a prompt yet, the canvas shows a few starter templates — Concept illustration, Labeled diagram, Cell biology, Anatomy, Chemistry model, and Historical scene — to give you a one-tap example to riff on.
Generating an Image
- Type a prompt in the box at the bottom, like "A clean, labeled illustration explaining how photosynthesis works, textbook style."
- (Optional) Pick a quality level, aspect ratio, and one or more models — see below.
- Hit the arrow button or press Enter.
Generated images appear as tiles on the canvas as they finish. Click any tile to open it full-screen, download it, or delete it.
Choosing a Quality Level
Quality is a simple toggle:
| Quality | What you get |
|---|---|
| Standard | Fast, everyday quality — available on every plan. |
| High | Sharper 2K output for when you want a crisper result. Available on paid plans; the free plan sees it with a lock badge you can tap to upgrade. |
If you pick High but the model you're generating with doesn't support it, the Playground automatically falls back to Standard for that image rather than failing.
Each tile on the canvas also shows which model produced it, so it's easy to tell your generations apart when you've used more than one at once.
Your quality pick is remembered for next time, so you don't have to keep picking it.
Picking an Aspect Ratio
Five aspect ratios are available — 1:1, 16:9, 9:16, 4:3, and 3:4 — and the Playground remembers your last pick. Use 16:9 for slide visuals, 9:16 for phone screens or stories, 1:1 for thumbnails, and 4:3 / 3:4 for handouts.
Choosing a Model (or Several at Once)
The Playground supports multiple image models and lets you generate with more than one model at the same time from a single prompt — useful when you're comparing styles, picking a favorite, or just don't yet know which model fits the look you want.
| Model | Notes |
|---|---|
| Nano Banana 2 Lite | The default. Google's fast, efficient image model for everyday study visuals. |
| GPT Image | Strong at text inside images, clean diagrams, and labeled visuals. |
| Grok Imagine | A more illustrative, photographic look — good when you want depth or atmosphere. |
| Nano Banana 2 | Google's higher-quality image model. Useful when you want a richer alternate pass. |
Tick more than one model and a single press of Generate creates one image per model from the same prompt. Your model selection is remembered between sessions.
Reference Images (Upload or Paste)
You can attach up to 4 reference images to a prompt to edit, remix, or blend them. Use the image button next to the prompt to upload from your device, or just paste an image straight into the prompt box. PNG, JPEG, and WebP are supported.
When you've added reference images, the prompt box switches to "Describe how to edit your image…" — instructions like "Make the background white and add labels for each organelle" are applied to your uploads. This works with Nano Banana 2 Lite, GPT Image, Grok Imagine, and Nano Banana 2.
Remove a reference with the small × on its thumbnail, and any image you generate can itself be used as a starting point — open it, download it, and re-attach it to a new prompt.
Background Generation
Images keep generating even if you refresh the page or leave the Playground entirely — when you come back, finished images are already on the canvas and any still-pending ones pick up where they left off. No need to sit and wait.
When a Prompt Can't Be Generated
If a prompt can't be turned into an image because it goes against our content guidelines, the Playground explains that up front so you can quickly rephrase, instead of failing silently a few seconds later. The same applies to prompts that ask for celebrities, copyrighted characters, or other restricted content.
Lifetime AI Creation Credits and Beta Limits
Each generation counts as one AI creation. On the free plan, those are part of your 1 lifetime AI creation credit and do not reset. While the Playground is in beta, it also has a separate small daily capacity cap; the remaining beta count is shown above the prompt box and resets daily. See Plans and Limits for the per-plan breakdown.
Playground vs. Image Generation in Chat
You can also generate images directly inside any AI Chat conversation — that's still the right tool when the image is part of a wider answer you're studying, when you want the AI to decide whether to draw something, or when you want a follow-up explanation alongside the image. See AI Image Generation for the chat-side flow.
Reach for the Playground when the image itself is the deliverable — you want a clean canvas, multiple variations from different models, more control over style and aspect ratio, and the ability to walk away while it generates.
Tips
- Be specific. "A labeled cross-section of a leaf with chloroplasts, stomata, and the xylem/phloem clearly marked, textbook style" gives a far sharper result than "a leaf."
- Generate with two models at once when you don't know which look you want — pick whichever you prefer and discard the other.
- Use a reference image of an existing diagram or hand-drawn sketch and ask the model to redraw it more cleanly.
- If a prompt is rejected for content reasons, rephrase or move it to a more educational angle — the Playground is tuned for study material.