Nano Banana Image to Video: Step-by-Step Guide

The most reliable way to get an AI video that looks exactly the way you planned is to design the first frame as a still image, then animate it. Nano Banana gives you precise, cheap control over the still: composition, character, product, light. A video model such as Veo, Seedance, Wan or Gemini Omni then adds motion and, on most models, sound. This guide walks through the whole workflow on NanoBananas, from the first-frame prompt to a finished multi-shot sequence, using one credit balance for both steps.

Model options, durations and credit prices in this guide are the ones the NanoBananas studio uses on November 9, 2026. The studio always shows the exact quote before you generate.

Why start from an image

• Control: you approve the composition, character and lighting before paying for video. A still costs 3 to 20 credits; an 8-second clip costs 24 to 360.
• Consistency: the same still, or edited versions of it, can open several clips, so a character or product looks the same across shots.
• Fewer failed clips: text-to-video has to invent the look and the motion at once. Image to video only has to invent the motion, which is where most failures happen anyway.
• Brand accuracy: logos, packaging and product colors are far easier to get right in a still, where you can edit and check them, than in a video prompt.

Step 1: Create a first frame that can move

Generate the still in the image studio. A good first frame is not just a good picture; it is a picture that leaves the video model somewhere to go.

• Use the video's aspect ratio. 16:9 for landscape, 9:16 for vertical social clips. Veo 3.1 and Gemini Omni accept only these two; Seedance and Wan 3.0 also accept 1:1, 4:3 and 3:4, and Seedance adds 21:9.
• Leave room for motion. A subject cramped against the frame edge has nowhere to walk; a product filling the whole frame cannot be orbited.
• Freeze an action mid-way. A barista mid-pour or a runner mid-stride gives the model an obvious next movement.
• Avoid tiny text and dense crowds. Both are hard to keep stable once things move.
• Make it at 2K. It is enough detail for 1080P video. 4K stills cost more and do not make the video sharper.

Use Nano Banana 2.1 or Pro for the final frame; draft compositions on Nano Banana 2 at 1K for 3 credits each. See how to write Nano Banana prompts for the prompt structure.

Step 2: Choose a video model

Every model below can animate a first frame in the studio. They differ in length, formats, sound and price.

• Short social clip with sound: Veo 3.1 Fast, or Lite to save credits.
• Longer single take: Seedance 2.5 (up to 30 s) or Wan 3.0 (up to 15 s at a low price per second).
• Square or 4:3 feed format: Seedance or Wan 3.0.
• Cheapest motion test: Veo 3.1 Lite or Seedance 2.0 Mini, 24 credits for 8 seconds at 720P.

For a deeper comparison, read Seedance vs Veo vs Wan and Veo 3.1 Quality vs Fast vs Lite.

Step 3: Describe the motion, not the scene

Open Image to Video and upload your still. The image already defines the scene, so the prompt should describe what moves, how the camera moves and what we hear. Re-describing the whole scene invites the model to change it.

• Name one main action and one camera move per clip. Two of each usually blur together.
• Use speed words: slowly, gently, suddenly, in one smooth motion.
• Describe sound when the model makes it: quiet café ambience, footsteps on gravel, no music.
• Keep the prompt in English; most video models follow English prompts most reliably.

Step 4: Test cheap, then render the final

1. Test the motion at 720P and the shortest length the model offers. On Veo 3.1 Lite a 4-second test costs 12 credits.
2. Check the clip against a short list: does the subject keep its face and shape, does the camera do what you asked, does anything warp at the edges, and is the sound right?
3. Fix one thing per run. Too little motion: add a clearer action. Too much: add subtle and static camera.
4. Render the final at the length and resolution you publish in, on the model whose test you liked.

Keep the image and the motion prompt fixed when you switch models, so you are comparing models rather than prompts.

Control the ending with start and end frames

Start-and-end-frame mode takes two stills and generates the motion between them. It is the best way to make reveals, before-and-after shots and transitions land exactly where you want. Veo 3.1 (all three tiers), Seedance 2.0, 2.5, Fast and Mini, Wan 3.0 and 2.7, and Gemini Omni 1.1 support it in the studio.

1. Create the start frame in Nano Banana.
2. Create the end frame by editing the start frame, not by writing a new prompt. For example: Keep the room, camera angle and furniture unchanged. Change the lighting from daytime to evening with the lamps switched on. This keeps the two frames close enough to connect.
3. Upload both frames and describe the transition: The light slowly fades from day to evening as the lamps turn on one by one. Static camera.

If the two frames differ too much, the model has to invent a large change in a few seconds, which is where warping and morphing appear. The photo editing guide covers the keep-and-change wording for the end frame.

Build a multi-shot sequence

Longer videos are made of several clips. The way to keep them consistent is to make every shot's first frame from the same source.

1. Make a hero still of the character or product.
2. Edit it into each shot's first frame: a close-up, a wide shot, a new location. Keep the face, outfit or packaging in the keep sentence each time.
3. Animate each frame with its own short motion prompt, on the same video model and settings.
4. Join the clips in an editor. To continue a shot exactly, take a still from the last frame of the previous clip and use it as the next first frame.

For more on keeping a character stable across images, see the character consistency guide.

Three image-to-video pairs to copy

Product hero shot

Portrait for social

Landscape establishing shot

Common problems and fixes

Frequently asked questions