AI image generator
Pick a model, set the resolution tier and aspect ratio, attach reference images if the result has to match something you already have, and generate.
Results
Generated content will appear here
What you can run here
Every card below opens a page that runs in this same workspace. Each one carries its own samples, limits and published prompts, so you can compare them before you commit a run.
AI video generator
Generate video with four Seedance tiers: Seedance 2.5 and the 2.0, 2.0 Fast and 2.0 Mini tiers. Compare duration, resolution and inputs, then create.
Image to prompt: read a reusable prompt from any photo
Turn any image into a reusable prompt: subject, style, lighting and lens, in the order each AI model reads it. Traceable examples and honest limits.
Doubao image watermark remover
Remove supported visible Doubao image watermarks free in your browser. No upload or signup; batch-process PNG, JPEG and WebP files locally.
Nano Banana Pro image generator and editor
Generate and edit with Nano Banana Pro, Google's Gemini 3 Pro Image model. Copy Google's published prompts, compare three tiers, and check the quote.
Nano Banana 2 image generator and editor
Generate and edit with Nano Banana 2, Google’s Flash-speed Gemini 3.1 model. Copy published prompts, compare it with Pro and 2 Lite, and check the quote.
Nano Banana 2.1 image generator and editor
Generate and edit with Nano Banana 2.1, Google's GA Flash-tier image model: open modes, limits, copyable prompts and the live quote before you send.
Nano Banana 2 Lite image generator and editor
Google’s fastest Nano Banana model: 1K images in about 4 seconds at $0.034 each. Official benchmarks, published prompts, and this workspace’s controls.
GPT Image 2 image generator
Generate and edit with GPT Image 2 on image 2.5: multi-script layouts, photoreal scenes and sizes up to 4K. See the official specs, limits and quote first.
GPT Image 2.5 AI Image Generator
Create with GPT Image 2.5 online. Generate from text, edit reference images, explore examples and prompt ideas, and check your quote before creating.
FLUX 3 Image generator and editor
Generate and edit with Black Forest Labs FLUX 3 Image: up to ten references, bounding-box layouts on a 0 to 1000 grid, and 4K output. Runs on this site.
One form, three jobs
This page is the image workspace itself. Text to image, editing a photo you already have and building from reference images all use the same form, and the model you pick decides which of the three a run can do.
- Text to image: describe the picture and generate it in one step.
- Editing: attach an existing photo, describe the change, and the run returns an edited version.
- Reference images: attach what has to survive the run, such as a product, a face or a layout, and say so in the prompt.
The seven models, side by side
All seven sit in the model picker. What changes between them in practice is the resolution tier you can ask for and how many reference images a single run accepts.
| Model | Resolution tiers | Reference images per run |
|---|---|---|
| Nano Banana 2.1 | 1K, 2K, 4K | up to 14 |
| GPT Image 2.5 | 1K, 2K, 4K | up to 10 |
| GPT Image 2 | 1K, 2K, 4K | up to 10 |
| Nano Banana Pro | 1K, 2K, 4K | up to 14 |
| Nano Banana 2 | 1K, 2K, 4K | up to 14 |
| Nano Banana 2 Lite | 1K | up to 5 |
| FLUX 3 Image | 1K, 2K, 4K | up to 10 |
How a run works
Five steps, in the order the form asks for them.
- Pick a model. The picker decides which aspect ratios, resolution tiers and reference-image limits are on offer.
- Write the prompt. Name what belongs in the frame before describing how it should look.
- Choose a format that matches the destination: 9:16 for a phone screen, 16:9 for a video thumbnail, 1:1 for a profile image.
- Check the price the workspace quotes, then send the run.
- Review the result at full size before you publish it. Hands, faces, small type and reflections are where a run usually needs a second attempt.
Prompt and settings do different jobs
The prompt decides what is in the frame. The settings decide how much of it the model can resolve.
- A named subject beats a pile of adjectives.
- Resolution is a ceiling rather than an upgrade. A 4K run leaves room for detail but will not invent detail the prompt never asked for.
- Reference images are for consistency, not collage. Say which element has to survive the run.
- Short labels survive better than sentences. Ask for a word rather than a paragraph when the image has to carry text.
What this page does not decide
Prices, parameters and the model line-up come from the live service. This page describes what the workspace offers today, and the workspace wins if the two ever disagree.
- The point cost of a run is quoted by the service before you send it, not estimated here.
- A model can be added or retired without this page being rewritten. The picker is the authority.
- The samples below are the publishers' own images. Nothing on this page grants a licence to reuse them.
Examples
Create an infographic that shows how to make elaichi chai
Nano Banana Pro returns a finished layout rather than a photograph: heading, numbered steps and a drawn illustration inside one image.
This aerial shot captures a dramatic, misty landscape, likely a valley or glen, characterized by rolling, verdant hills and a winding river or loch. The photography style leans towards a moody and atmospheric aesthetic, emphasizing the grandeur and isolation of nature. The camera angle is high, looking down into the valley, providing a sweeping panoramic view that highlights the immense scale of the surroundings. The dominant colors are various shades of deep green, ranging from lush emerald in the foreground fields to darker, more muted tones on the distant mountains.
Nano Banana 2 holds a long, specific photographic brief — camera angle, mood and dominant colours — inside a single prompt.
GPT Image 2 has to make the type legible: this poster works as branding, packaging direction and product photograph at once.
Questions
Which models can I use here?
Seven. Nano Banana 2.1, Nano Banana Pro, Nano Banana 2 and Nano Banana 2 Lite from Google, GPT Image 2.5 and GPT Image 2 from OpenAI, and FLUX 3 Image from Black Forest Labs. All seven are in the model picker on this page.
Can I edit a photo I already have?
Yes. Attach the photo and describe the change. Every model here accepts reference images, and a single run takes between five and fourteen of them depending on which model you picked.
Which model should I start with?
Nano Banana Pro is the picker's default. Move to another model when you want a different set of aspect ratios, a lighter run, or the specific look you have already seen on that model's page.
How much does a run cost?
The workspace quotes the price before you send it. The amount is set by the live service rather than by this page, so the figure in the form is the figure to trust.
What are reference images actually for?
Keeping something from one image in the next one: a product, a face, a logo, a colour or a layout. Attach the image and name the part that has to survive the run.
Do the samples on this page come from this site?
No. Each sample below is the model publisher's own launch image, shown with the prompt they published beside it. They show what each model family produces, and none of them was generated here.
Why is there no price table on this page?
Because a run's cost moves with the model, the resolution tier and the reference images included, and a written table would go stale without anyone noticing. The form quotes the live figure instead.