Nano Banana Video: How to Animate Nano Banana Images
October 27, 2026
Nano Banana can't make video. It's Google's image model, and every version of it outputs still images. When people say "Nano Banana video," they mean a two-step job: make the picture with Nano Banana, then hand it to one of Google's video models to animate it. In October 2026 those video models are Gemini Omni (which has replaced Veo in the Gemini app) and Veo 3.1 (still available in Google Flow and on Google Cloud).
That split helps. A still is easier to get right than a clip, so you lock the character, outfit and lighting in Nano Banana first, then spend your video attempts on motion only.
Here's the whole thing in one table:
| Step | Google tool | What it does |
|---|---|---|
| 1. Make the frame | Nano Banana (in the Gemini app or Flow) | Generates or edits the still image: your character, scene, or product shot |
| 2. Animate it | Gemini Omni or Veo 3.1 | Turns that image into a short clip, as a first frame, a start-and-end pair, or a reference |
| 3. Extend or edit | Gemini Omni (extension via the Gemini API; coming soon for Omni in Flow) | Continues the scene or changes it through follow-up prompts |
Below: which model does what now, the workflow in the Gemini app and Flow, and prompts for each step.
Can Nano Banana create video?
No. Google's Nano Banana developer docs describe Nano Banana as the name for Gemini's native image generation. The family covers several models (Nano Banana 2.1, Nano Banana Pro, Nano Banana 2 Lite and older ones), and all of them generate and edit images. None returns a video file.
The confusion is fair, because Google pitches its video model through the image one. Google DeepMind's Gemini Omni page puts it this way: "Think of Gemini Omni like Nano Banana, but for video." Same conversational style, same idea of editing by talking to it, different output.
So the honest answer to "nano banana create video" is: Nano Banana creates the frames, and a video model moves them. If you only want stills, our GPT Image 2 vs Nano Banana guide explains which Nano Banana model is which.
Google's video models in October 2026: Gemini Omni and Veo
This changed recently, and many guides still say "use Veo in Gemini." Here's the current state, from Google's own pages.
Gemini Omni replaced Veo in the Gemini app. Google's Gemini video generation page says Gemini Omni is the latest video generation and editing model and takes Veo's place in the app. The current version is Gemini Omni 1.1 Flash.
Veo 3.1 previews are gone from the Gemini API. Google's video generation docs say the Veo 3.1 preview models were shut down on October 22, 2026, and point developers to Gemini Omni Flash or to the GA Veo models on the Gemini Enterprise Agent Platform.
Veo 3.1 still exists elsewhere. It's listed in Google Flow (as Lite, Fast and Quality), and Google Cloud's Veo 3.1 model page lists veo-3.1-generate-001 as GA, with image-to-video, first-and-last-frame and reference image support.
| Gemini Omni 1.1 Flash | Veo 3.1 | |
|---|---|---|
| Gemini app | Yes, it's the app's video model now | Replaced by Omni |
| Google Flow | Yes | Yes: Lite, Fast and Quality |
| Gemini API | Yes (gemini-omni-1.1-flash) | Preview models shut down October 22, 2026 |
| Gemini Enterprise Agent Platform | Yes | Yes, GA |
| Clip lengths in Flow | 4, 6, 8 or 10 seconds | 4, 6 or 8 seconds |
| Editing by follow-up prompt | Yes | No video-to-video edit in Flow |
Two Omni features matter most for the Nano Banana workflow. Google's Omni 1.1 Flash announcement (August 27, 2026) added first and last frame control, so you can set a start image and an end image and let the model fill the motion in between, and scene extension in 10-second steps up to 40 seconds total (in the Gemini API; Flow lists extension for Omni as coming soon). It also added 360p drafts for cheaper test runs, with upscaling to 1080p or 4K for the keeper.
The Nano Banana to video workflow, step by step
The pattern is the one Google's own Omni prompt guide suggests: reference images for Omni "can be real photos or images made with Nano Banana." You build the look as stills, then animate.
Step 1: Make your keyframe or character in Nano Banana
In the Gemini app, open the tools menu, pick Create images, and prompt. Google's Gemini image generation page lists Nano Banana 2 as the free model, with Nano Banana 2.1 on a Google AI plan.
Decide which of three things you're making:
- A start frame. The exact first image of your clip, composed the way you want the shot to open.
- A start frame and an end frame. Two images of the same scene, showing where the motion begins and where it lands.
- Ingredients. Separate clean images of a character, a product, or a location that the video model combines into a new scene.
Make the frame in the aspect ratio you'll publish in. Both Omni and Veo support landscape (16:9) and portrait (9:16), so a 9:16 still saves you from cropping later.
Step 2: Animate it in the Gemini app
This is the fastest path. Per Google's video generation page, Gemini Omni in the app can turn up to five photos into a video, generates clips up to 10 seconds with native audio, and needs a Google AI subscription. Features vary by plan tier and region.
- Download your Nano Banana image (or keep it in the same chat).
- Start a video request and attach the image.
- Describe the motion, the camera, and the sound. Don't re-describe what's already in the picture.
- Review the clip, then ask for changes in plain language: "slower camera," "keep everything the same but make it night."
Step 3: Or animate it in Google Flow for more control
Google Flow is Google's AI creative studio for video and images, and it's where you get the frame controls. Flow's models page lists Nano Banana Pro, Nano Banana 2.1 and Nano Banana 2 Lite for images, plus Veo 3.1 and Gemini Omni Flash 1.1 for video, so you can make the still and the clip in one project.
According to Flow's help page on creating videos:
- Frames: click the model name, then Video, then Frames. Drag an image onto "Add start frame," optionally another onto "Add end frame," and describe the action between them. Omni's start-and-end frames currently work on Flow for web only.
- Ingredients: click the model name, then Video, then Ingredients. Drag in reference images, or type @ to pull in assets from your project by name. Your prompt should say how each ingredient appears.
Flow's tip for ingredients: plain backgrounds for subject and product references, and a consistent look across all of them. That's what Nano Banana is good at producing.
Step 4: Extend and edit
What you can do next depends on where you made the clip.
- In the Gemini app: you don't start over when something's off. Ask for the change in a follow-up turn and Omni keeps the rest.
- In the Gemini API: Omni can also extend a clip. Describe what happens next and it generates a continuation. Google's Omni API docs note that extension only appends to the end of a clip, never the middle.
- In Google Flow: Flow's models page lists Extend videos for Gemini Omni Flash 1.1 as coming soon, so the API's extension doesn't carry over yet. You can still edit an Omni clip of up to 10 seconds with video-to-video editing. To make it longer, chain clips: use your Nano Banana end frame as the start frame of a new Frames generation. Flow's native Extend currently works with Veo 3.1 Lite, on 8-second clips only.
Nano Banana video prompts for each step
Swap in your own subject. The structure follows Google's prompt guide, which builds a video prompt from shot framing and motion, style, lighting, location and action.
Prompt 1: A character sheet in Nano Banana
Use this when the same person or mascot needs to appear in several clips.
Character reference sheet of a young woman with a short black bob, round
glasses and a mustard-yellow raincoat. Show her front, three-quarter and
side views on a plain light grey background. Even studio lighting, neutral
expression, full body visible. Clean, consistent illustration style.
Prompt 2: A start frame in Nano Banana
Vertical 9:16 photo. The same woman from the reference stands at a rainy
Tokyo crosswalk at dusk, holding a clear umbrella, neon signs reflected in
wet pavement. Medium shot, eye level, shallow depth of field. Cinematic,
soft teal and orange tones.
Prompt 3: An end frame in Nano Banana
Upload your start frame and edit it, so everything except the change stays put:
Keep the scene, framing, lighting and character exactly the same. Change
only the moment: the light has turned green, she has stepped into the
crosswalk mid-stride, and she is smiling at someone off-camera to the left.
Prompt 4: Animate the start frame
Google's Omni docs say vague prompts like "make it move" give weaker results than detailed descriptions of camera movement, subject motion and environment. So spell all three out:
Slow push-in on the woman as the light changes. She lowers the umbrella
slightly, looks up and smiles. Rain keeps falling, neon reflections ripple
in the puddles, a cyclist passes in the background. Ambient city rain and
distant traffic. One continuous shot.
Prompt 5: Start and end frames
Smooth transition from the first frame to the last frame. She steps off the
curb into the crosswalk; the camera tracks alongside her at walking pace.
Umbrella sways with her stride. Keep her face, outfit and the lighting
consistent the whole time.
Prompt 6: Ingredients
Attach your character sheet and a separate Nano Banana image of a café interior:
The woman from the character sheet walks into the café, shakes rain off
her umbrella and waves at the barista. Handheld camera, medium shot, warm
interior light against the blue rain outside. Her coat and glasses match
the reference exactly.
For more still-image prompts to feed this workflow, see our Nano Banana prompts collection. Trend formats work too: a Gemini Polaroid image makes a natural start frame for a short "photo comes to life" clip.
Why clips break, and the fix for each
| Problem | Likely cause | Fix |
|---|---|---|
| Face or outfit changes mid-clip | The model is inventing details the frame didn't pin down | Add a character sheet as an ingredient, and say "keep her face and outfit exactly the same" |
| Nothing much happens | Motion prompt was too short | Describe camera move, subject action and background motion separately |
| Clip crops awkwardly for Reels or Shorts | Still was made in the wrong aspect ratio | Generate the Nano Banana frame in 9:16 from the start |
| End frame doesn't match | You regenerated the end frame from scratch | Make it by editing the start frame, changing one thing |
| Text in the frame wobbles | Small lettering is hard to keep stable in motion | Keep text large and simple, or add it in an editor afterwards |
Test at low resolution first. Omni's 360p drafts exist for this, and 1080p and 4K are upscaled outputs you pick once the motion is right.
All of this comes with Google's provenance marking. The Gemini video page says every video generated in the Gemini app carries an invisible SynthID watermark, and the Omni API docs say the same for API output.
FAQ
Can Nano Banana make videos?
No. Nano Banana is Google's image generation model, and it only outputs images. To get video, you animate a Nano Banana image with Gemini Omni or Veo 3.1.
Is Veo still available?
Yes, but not everywhere. Veo 3.1 is in Google Flow and is GA on the Gemini Enterprise Agent Platform. In the Gemini app, Gemini Omni has replaced it, and the Veo 3.1 preview models in the Gemini API were shut down on October 22, 2026.
What is Gemini Omni?
Google's model for generating and editing video from any mix of text, images, video and audio. It's the video model in the Gemini app now, and it's also in Flow, Google AI Studio and the Gemini API.
Is Nano Banana video free?
Nano Banana image generation has a free model in the Gemini app. Video in the Gemini app is different: Google's video page says Gemini Omni requires a Google AI subscription, and features vary by tier and region. Check Google's pages for the current plans before you commit.
How long can a Nano Banana video be?
A single generation tops out at 10 seconds with Omni and 8 seconds with Veo 3.1 in Flow. Through the Gemini API, Omni's scene extension keeps adding 10-second segments up to 40 seconds total. In Flow, extension for Omni is listed as coming soon, so chain clips with start and end frames, or use Veo 3.1 Lite's Extend on 8-second clips.
Can I keep the same character across several clips?
Yes. Make one clean character sheet in Nano Banana, use it as a reference in every clip, and repeat the key details (hair, clothes, accessories) in each prompt.
Key takeaways
- Nano Banana makes images only. "Nano Banana video" means animating its stills with a video model.
- Gemini Omni replaced Veo in the Gemini app. Veo 3.1 remains in Flow and on Google Cloud.
- Lock the look in stills first, then use Flow's Frames and Ingredients for control.
- Prompt the motion, not the picture: camera move, subject action, background motion and sound.
If you end up doing this every week, frame prompt, edit pass, motion prompt, extension, the steps you've tuned are worth keeping as a workflow you can run again rather than retyping. Taku is an AI-native desktop workspace where you can mirror an AI app or workflow someone already got working, run it, and save your edited version as your own. Taku is in Beta, and the Mac app is available now.