Sharper detail, clearer text, more consistent edits
Nano Banana 2.1 AI Image Generator
Create and edit images with Nano Banana 2.1 right here. Describe an idea or bring up to 10 reference images, choose 1K, 2K or 4K output, and get crisp, legible results — posters, product mockups, portraits and infographics — in seconds.
Leave empty to generate from text. Upload photos to guide identity, product shape or style.
Your generated image appears here
Skirt-move studio · experimental
1 pick · 1 charge · cloud render
One good move, already made — add your place and your word
The turn comes from a reference take, not from a sentence you have to get right. You bring the setting and the word on the skirt, and you pick how long and how large. This is cloud rendering, not a timeline editor: one request, one charge, 1–3 minutes.
Up to 5 Chinese characters or 10 letters.
0 / 10Advanced · prompt 0
Filled from the template. Edit it freely — the turn still follows the reference clip you picked.
Pricing is loading — reload if it stays blank.
What happens when you generate 1 request · 1 charge
Both inputs are optional. Leave the background empty and she stands on the reference clip's own set; leave the word empty and nothing is lettered on the skirt.
One request runs when you click Generate: the clip you picked is sent as the reference take, together with any photo, and the model renders the whole thing in one pass (1–3 minutes). It is billed once, by the length and resolution you chose.
Upload a background you have the right to use — no portrait is collected here. Uploads are moderated, and photos containing real people, celebrities or anyone under 18 are refused. Every clip carries a visible AI-generated label.
5s · 768P · 9:16 · MiniMax H3 reference
4–15s
You set the length — the master's own take is the default
9:16
Vertical by default, 768P or 2K, made for TikTok, Reels and Shorts
1
Request per clip — one charge, priced by length and resolution
10
Width units on the skirt: 5 Chinese characters or 10 letters
What the master already carries
A longer prompt does not make the spin better — it makes it luckier. A master is the take itself, already performed, and the clip is generated with that take as its reference, so four things stop drifting at random.
The turn
How far she spins and the moment the fabric lifts are taken from the reference clip, not from a sentence you have to get right.
The camera
The camera path is referenced from the take, so the reveal lands near the same beat instead of arriving wherever a fresh draw puts it.
The length
The take sets the starting length — 8 seconds on the live one — and you can trim it to 4 or stretch it to 15 from the slider.
The fabric
The scene asks for heavy cloth that carries momentum, so the hem behaves like a skirt — not a flat cut-out swung around.
How a Skirt Swirl clip gets made
Three steps, and nothing to timeline.
Pick a move
Four masters are listed above; one is live today and three are marked Soon. The number on each card is the length of that take.
Add your place and your word
A background photo is optional — skip it and she stands on the master's own set. The word is optional too — skip it and nothing is lettered on the fabric.
Let the cloud render it
One request goes out with the take you picked as its reference, plus your photo if you brought one, and the model renders the whole clip in a single pass (1–3 minutes). A failed run never spends credits.
No portrait is collected here, so the woman on the skirt is drawn fresh on every run — a different face each time. The take fixes the move; it does not fix who performs it.
Optional input · 1 of 2
A place to stand
Upload a photo of somewhere and she is set inside it, keeping that place's light and perspective. Skip it and she stands on the reference clip's own set — the Jimo gate tower on the live take.
- Under 30 MB · JPEG, PNG or WebP
- Long edge at least 256 px — a phone photo is easily past it
- Long edge under 5760 px — a normal photo, not a stitched panorama
Optional input · 2 of 2
A word on the fabric
A city, a name, a short wish — the word goes across the opened skirt in gold. Skip it and nothing is lettered, which is the cleanest way to watch just the turn.
- 10 width units: 5 Chinese characters or 10 letters — a hanzi counts double
- Chinese characters, letters, numbers and spaces; nothing else gets through
- Two to four characters is where the letters stay legible mid-spin
What lands in your file
One vertical clip
Vertical 9:16 by default, 768P or 2K, 4–15 seconds — you set the frame and the length on the left, and it is sized for the phone feeds it is meant for.
A label you cannot lose
Every clip carries a visible AI-generated label and an identifier inside the file. That is what TikTok, Instagram and YouTube policies and the EU AI Act ask for.
Keep the label on
Posting it with the label is the rule, not a watermark to crop out — removing it is what breaks those policies. What this site is.
One move. Your place. Your word.
If you have seen the clip — a dancer in an enormous skirt, one slow turn, and the fabric lifts into a fan that fills the frame with a word in gold across it: that is this. The dress with metres of hem, a dancer who can actually spin, and a location worth naming are the part nobody has at home. The master already has them.
Made with Nano Banana 2.1
Every image below was created from a short prompt or a single reference photo — collectible figures, product scenes, portraits and stylized art, rendered with sharp edges and legible detail.
Built for realistic, legible, consistent images
Google's latest image update improves realism, prompt following and text rendering — the details that make an image usable, not just pretty. Explore the capabilities below.
Real-world Knowledge
Grounded reasoning helps the model deliver precise, detailed results — annotate a picture, represent data as an infographic, or turn a handwritten note into a clean diagram.
Clear Text Generation
Sharp, legible lettering for posters, diagrams and product mockups. Describe the wording and where it belongs, and check spelling and alignment before you publish.
Perfect Style Transfer
Capture and apply an artistic style — oil painting, watercolor, illustration or photography — and keep it consistent across a series of creations.
4K Ultra Resolution
Choose 1K for a quick concept, 2K for a detailed review, or 4K for larger layouts — every detail crisp enough for print and commercial use.
Image to Image: edit with a reference
Upload a photo and describe the change — restyle a portrait, swap a background, or turn a concept into a product mockup while keeping the details that matter.
- Guide the result: A clear reference image locks identity, product shape or composition so edits stay consistent.
- Compare models: Nano Banana 2.1, Nano Banana Pro, GPT Image-2 and Flux-2 Pro on one platform.
Text to Image: from a line to a layout
Describe the subject, camera angle, lighting and exact wording, and the model renders an original image in seconds — no canvas, no stock search.
- Type your prompt: Put required wording in quotes and give it a position in the layout.
- Pick a format: Choose an aspect ratio and 1K, 2K or 4K resolution before you generate.
The trends everyone is remaking
Bulin 47 AI Freestyle
Put your photo on the viral red-lit doorway set — crowd, beat and all. See how the trend works and make your take.
Hotel Lobby AI Duet
Two photos become a 3D animated duet on the glowing orange studio set — one hanging mic, shared by two.
Skirt Swirl AI Dance
One turn and a giant skirt opens into a fan that fills the frame, a word in gold across it. Two photos put you on that set.
AI Video Effects
Hundreds of one-tap templates that turn a photo into a short clip — pick an effect and generate.
How to Create with Nano Banana 2.1
Three simple steps from a creative brief to a finished image — no design skills and no software required
Choose Model, Format & Reference
The workbench defaults to Nano Banana 2.1. Select text-to-image for a new composition or add a reference image for an edit, then pick an aspect ratio and 1K, 2K or 4K resolution.
Specify the Scene and Text
Describe the subject, framing, lighting and background. Put any required wording in quotes and give it a position. For an edit, say what should change and what should stay.
Generate, Inspect, Refine
Check the credit cost shown before you generate, then inspect faces, product details and lettering. Make a focused follow-up edit and download the version that fits your layout.
Prompts that get you closer, faster
A precise prompt sets the ceiling for the result. These checks avoid the common text, identity and composition problems before you generate.
Write the whole scene
Subject, setting, framing, lighting and format. "A matte white skincare bottle on pale stone, soft window light from the left, centered 4:5 composition" beats a single word.
Quote exact wording
Put any text you need in quotation marks and say where it belongs. Then check spelling, alignment and small details before you publish.
Edit in focused steps
For an edit, name the one change you want and the details to keep. Small, specific follow-ups hold the result steady across turns.
FAQ
Sharper detail, clearer text.
One prompt or one reference image in, a print-ready visual out. Put your idea into Nano Banana 2.1 and see the result in seconds.