Nano Banana Pro is Google DeepMind's image model built on Gemini 3 Pro. Most of its strengths come from the fact that it understands instructions, not just keywords. You can describe a scene the way you'd brief a photographer, ask it to change one thing in an existing image, or give it several references and have it combine them. Text inside images comes out legible, it keeps up to five characters consistent, and it outputs at 1K, 2K or 4K.
This guide covers a prompt structure that works, eight example prompts you can copy, and how to use the result as a keyframe for Seedance 2.5 image-to-video.
Nano Banana vs Nano Banana Pro
There are a few versions of the model. The original Nano Banana is fast and cheap: good for mood boards, sticker packs and lots of variations. Nano Banana Pro is the one to use when you need precise text rendering, multiple references, character consistency or 4K output. Everything below works on both, but the editing and typography examples benefit most from Pro.
The prompt structure that works
Write in full sentences, roughly in this order:
- Subject: who or what, with the specific details that matter.
- Setting and composition: where it is, framing, what's in the foreground and background.
- Camera and lens: shot size, angle, depth of field.
- Lighting and color: time of day, light source, grade.
- Style and finish: photographic, illustrated, 3D, etc.
- Text (if any): put the exact words in quotes and say where they go.
- Constraints: what must stay the same or must not appear.
For edits, flip it around: say what to keep first, then the single change you want. Small, specific edits work much better than "make it better".
8 example prompts you can copy
1. Product hero shot (text-to-image)
A matte black ceramic coffee mug on a light oak table, morning sunlight coming from the left through a window, soft shadows, a thin wisp of steam rising from the cup. 50mm lens, eye level, shallow depth of field with the background softly blurred. Clean, warm commercial photography, neutral color grade. No text, no logos.
2. Poster with readable typography
A minimalist travel poster for Kyoto in autumn: a single red maple branch in the upper right corner, a pagoda silhouette at the bottom, cream paper texture background. The headline "KYOTO" in bold geometric sans-serif at the top center, and the smaller line "Autumn 2026" below it. Flat illustration style, limited palette of red, charcoal and cream.
Quoting the exact text and saying where it goes is the main thing that gets clean lettering out of Pro.
3. Consistent character, new scene
Upload a reference image of your character, then:
Keep the same woman from the reference image: same face, hairstyle and green raincoat. Place her on a rainy Tokyo street at night, holding a clear umbrella, neon signs reflecting on the wet pavement. Medium shot, cinematic lighting, slight teal-and-orange grade.
4. Single-change edit (image-to-image)
Keep everything in this photo exactly the same, but change the sky to a soft pink sunset with scattered clouds, and adjust the lighting on the building to match the warm sunset light.
5. Multi-image fusion
Upload a product photo, a background photo and a style reference:
Place the sneaker from image 1 on the concrete steps from image 2, lit to match the scene. Use the color palette and high-contrast mood of image 3. Low angle, the sneaker sharp in the foreground, background slightly out of focus.
6. Infographic
A clean vertical infographic titled "How Coffee Is Made" with four numbered steps: "1. Harvest", "2. Dry", "3. Roast", "4. Brew". Each step has a simple flat icon and a one-line label. White background, brown and orange accent colors, modern sans-serif type, generous spacing.
7. Character sheet for animation
A character turnaround sheet of a small friendly robot with a round body, one large blue eye and stubby arms: front view, side view, back view and three-quarter view in a row, plain light gray background, consistent proportions and colors in every view, soft 3D render style.
8. Cinematic keyframe (made for video)
Wide cinematic shot of a lone hiker in a red jacket standing on a mountain ridge at sunrise, low clouds rolling through the valley below, golden light on the peaks. The hiker is small in frame on the left third, facing right toward open space. 21:9 composition, natural colors, high detail, no text.
The next section explains why this one is written the way it is.
Turning your image into a Seedance 2.5 video
A strong still makes a strong first frame. Here's the workflow I use:
1. Compose for motion. Leave room in the direction the subject will move or the camera will travel. In prompt #8 the hiker faces open space on the right, so a slow pan or push-in has somewhere to go. Avoid heavy text in the keyframe. Video models can warp lettering once it starts moving.
2. Match the aspect ratio. Generate the image at the ratio you want for the video. Nano Banana Pro supports 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3 and 21:9. Seedance 2.5 supports 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 and adaptive. Seedance renders at up to 720p, so a 1K or 2K image is plenty for a keyframe. Save 4K for print or stills.
3. Write a motion prompt, not a picture prompt. The image already defines the look. Describe what changes:
Slow push-in toward the hiker as the clouds drift through the valley, the sun rises slightly and the light warms on the peaks, the jacket moves gently in the wind. Soft wind ambience. Steady, cinematic camera.
4. Choose length and resolution. Seedance 2.5 lets you pick any whole number of seconds from 4 to 30 at 480p or 720p. Draft at 480p for 4–6 seconds, then render the version you like at 720p.
5. Use First & Last Frames for transformations. Want the mug from prompt #1 to go from empty to full, or a room to go from day to night? Make both states with Nano Banana Pro (use prompt #4's "keep everything the same" pattern for the second one), then upload them as first and last frame in Seedance 2.5. The model fills in the motion between them.
6. Chain shots. Turn on "Return Last Frame" in Seedance 2.5 and you can reuse the final frame as the starting image for the next clip.
If you want 2K video or native stereo sound on a short 4–15 second shot, the same keyframe also works with MiniMax H3.
Cost and getting started
On Veemo, a Nano Banana Pro image costs 30 credits at 1K or 2K and 40 credits at 4K. Seedance 2.5 video is billed per second: 22 credits/s at 480p and 47 credits/s at 720p. New accounts get 16 free credits after verification (valid 30 days), enough to try a short 4-second MiniMax H3 clip. Subscriptions start at $29.90/month (598 credits), or you can buy one-time credit packs that never expire, starting at $9.90.
Common mistakes
- Keyword soup. "Mug, 8k, ultra, masterpiece" does less than one clear sentence.
- Too many edits at once. Change one thing per pass, then iterate.
- Unquoted text. If you want specific words in the image, put them in quotes.
- Busy keyframes. Clutter in the still becomes flicker in the video.
FAQ
Can I use Nano Banana Pro images commercially?
Yes, they can be used for marketing, advertising and products. Generated images include a SynthID watermark that marks them as AI-generated.
How many reference images can I use?
Nano Banana Pro is built for multi-image fusion (Google lists up to 14 references). The Veemo generator shows the upload limit for the mode you're in.
What resolution should I use for a video keyframe?
1K or 2K is enough, because Seedance 2.5 outputs at up to 720p.
Do I need to describe the image again in the video prompt?
No. Describe the motion, camera and sound. The image already covers the look.
Is there a free way to try this?
Yes. New Veemo accounts get 16 free credits after verification.
Top comments (0)