Covers and banners
Wide 16:9 and 21:9 formats for an article header, a video thumbnail or a banner.
CAPABILITY
AI image generation from text
Describe a picture in ordinary words and get a finished file. Ten aspect ratios, up to four variants per run, and a choice of model for the job: one holds photorealism better, another handles drawn graphics.
- 10
- aspect ratios
- up to 4
- variants per run
- 4K
- maximum resolution
How it works
-
Describe the picture
Plain text, no special syntax. If the description comes out short, the built-in prompt enhancer expands it into a detailed one before generation.
-
Set the parameters
Aspect ratio, quality tier, number of variants and a ready-made style: photoreal, cinematic, anime, illustration, 3D toon, editorial.
-
Start it and get on with your day
Generation runs in the background, so you can switch tabs or minimise the browser. Each variant appears the moment it is ready, without waiting for the rest.
-
Collect the result
Finished images land in your personal gallery together with the prompt and dimensions. From there you can download them, several at once if you like, or get a public link.
What it does
- Input
- A text description of the picture
- Aspect ratios
- 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9
- Resolution
- Each model has its own list of sizes; the top models go up to 4K, the rest have a lower ceiling
- Variants per run
- Up to four, each shown as soon as it is ready
- File formats
- PNG, JPEG, WEBP; some models return vector SVG
- Styles
- Photoreal, cinematic, anime, illustration, 3D toon, editorial and others
- Reference images
- Some models accept sample images and work from them
- Available in
- The dedicated generation screen, Studio and the regular chat
What people use it for
Vertical content
9:16 and 4:5 for vertical video formats and feeds where a horizontal image gets cropped.
Illustrations for articles
Four variants per run let you pick the right one instead of settling for the first.
Graphics for presentations
Styles hold a single visual language: the same line added to the prompt gives a consistent set of images rather than a mismatched one.
Models for this task
Full catalog
Image
Google Nano Banana 2 Lite
Lighter Nano Banana 2 image model for faster generation and editing workflows
Free: 3 per day
Image $0.042
Image
Seedream 5.0 Flash
Fast, cost-efficient ByteDance Seedream 5.0 tier: image generation and editing at 1K/2K with up to 10 reference images
Free: 3 per day
Image $0.023
Image
GPT Image 2
OpenAI image model: precise prompt following, readable text inside images, layout-aware posters, packaging and diagrams
Image $0.08
Image
GPT-Image-2.5 Flare
OpenAI's fastest image model: higher quality than the previous generation at half the latency
Image $0.10
Image
GPT-Image-2.5 Sunburst
OpenAI's most capable image model for complex layouts, reference fidelity and iterative editing
Image $0.18
Image
Grok Imagine Image 2.0
Next-generation Grok image model for text-to-image and prompt-guided editing at 1K and 2K
Image $0.06
Image
Kling IMAGE 3.0
Kling image model: native 2K output, realistic textures and materials, image-to-image from a reference
Image $0.034
Image
Krea 2 Large
Larger Krea 2 variant for rawer, more flexible outputs with stronger photorealism and weighted reference control
Image $0.076
Image
Krea 2 Medium
Faster Krea 2 variant for stable, consistent generation with controllable prompt strength and weighted reference guidance
Image $0.036
Image
Krea 2 Turbo
Krea 2 Turbo — the speed tier of Krea 2: aesthetic images in seconds, style references and moodboards, 8 aspect ratios.
Image $0.052
Image
Nano Banana 2.1
Google's newest Nano Banana: image generation and editing at 1K/2K/4K with up to 14 reference images
Image $0.042
Image
Qwen-Image-3.0
Alibaba image model balancing quality and speed: generation, editing, multilingual typography, up to 3 references
Image $0.036
What else it does
The things people usually discover only once they start working.
- A negative prompt is available on models that support it.
- Several variants from one run are available in Studio and on the generation screen.
- A non-standard size is snapped to the nearest one the selected model supports.
Frequently asked questions
How much does one image cost?
It depends on the model, and the figure is shown on the run button before you press it; the charge matches that amount. Full per-model pricing is on the pricing page.
What exactly am I charged for?
For a successful result: money is taken only for a finished image, and the amount matches the price shown on the run button.
Can I get several variants from one description?
Yes, up to four per run. Each one appears in the interface as soon as it is ready, so you see the first variant without waiting for the last.
What format does the file come in?
PNG, JPEG or WEBP depending on the model; some models can return vector SVG.
Can I generate an image inside a conversation?
Yes. Ask a text model to draw something: it calls the generator itself and returns the image in the chat. When you need several variants from one run, use the generation screen or Studio.
Where do I find what I generated?
Everything is saved to your personal gallery together with the prompt and dimensions. From there you can download the file, several at once if needed, or get a public link to send to someone else.
Related capabilities
Try it on your own task
A starting balance is already on your account, no card required. You pay only for what you generate.