Nano Banana 2.1: Google’s Image Model, Used for Video
A still from Nano Banana 2.1 is the most controllable frame you can get before a clip moves. Here is what the model does, what it costs, and how to turn its images into video.
Nano Banana 2.1 is Google’s image generation and editing model, released in general availability on 6 October 2026. It sits in the Gemini Flash family and is built for the job most teams actually have: producing a high-quality still — a keyframe, a product shot, a character sheet — and editing it precisely. It takes text, images, video and PDFs as input, and it outputs images and text; it does not output video.
That last point is the one that matters for video creators. Nano Banana makes the frame; it does not move it. The practical workflow, and the one this page is built around, is to generate the still with Nano Banana 2.1 and then animate it with a separate image-to-video model. OpenFilmo does not run Nano Banana itself — it is the step that comes after the image is made.
What it does
An image model built for the frame before the motion.
Text and image to image
Describe a scene or feed in references. Multi-image fusion accepts up to 14 reference images, so a character or product can be established once and reused.
Character and object consistency
The model is documented as holding up to four characters and up to ten objects consistent across a set of images — the same problem a storyboard or a keyframe sequence needs solved.
Mask-based editing
Mark a region and change only that area across turns, instead of regenerating the whole frame. This is what makes an iteration loop usable rather than a restart every time.
Wide ratios and 4K with provenance
Fourteen aspect ratios include 1:4, 4:1, 1:8 and 8:1 panoramic frames, fixed at 2K and 4K. Every generated image carries a SynthID watermark.
From a still to a clip
The image-to-video workflow, step by step.
Generate the still in Nano Banana 2.1
Pick the ratio and 1K, 2K or 4K, and describe the frame as a still. Work at the resolution the final delivery needs so composition is settled before motion is added.
Send the still to an image-to-video tool
OpenFilmo’s image-to-video generator takes your Nano Banana still as the first frame. There is no re-upload to a separate account — the image is already a file you control.
Describe only the motion
Appearance is already fixed by the still. Spend the prompt on what moves and how the camera moves: a slow orbit, hair in the wind, steam rising.
Render, upscale, download
Generate the clip, then upscale to 4K on the final cut. OpenFilmo’s Free plan carries no watermark, so a test cycle costs nothing before you commit.
What people make with it
The jobs this actually gets used for.
Storyboards and keyframes
A character sheet plus a few props can anchor every frame in a sequence. Multi-image fusion is the mechanism that keeps those frames on-model.
Product and ecommerce stills
Generate the hero shot, adjust it locally, then animate the same asset for a listing video without a reshoot or a second tool.
Anime and character sheets
Consistent characters across images are exactly what stylised video needs, because character drift is what breaks a sequence.
Thumbnails, posters and covers
Where the still is the deliverable and the point is a clear subject at small sizes rather than detail that disappears.
Why an image model is part of a video workflow
The still is the control surface. Motion is what you add afterwards.
Image-to-video is the most predictable way to get usable AI video, because you decide the frame and the model only supplies the motion. A still you have edited and approved is a far better starting point than a prompt hoping for a good first frame, and Nano Banana 2.1 is built to make that still worth approving — sharp at 4K, consistent across references, editable by region.
Consistency is the hard part of any multi-shot piece, and it is where stills earn their place. Holding up to four characters and ten objects across a set of images means the keyframes of a sequence can actually match, which is what lets the later animation read as one production instead of a montage of near-misses.
The honest framing is that this is a two-step pipeline, not one model doing everything. Nano Banana is the best-in-class image step on Google’s pricing; the video step is a different tool with different strengths. Routing each step to what it is good at beats forcing one model to span both, and it keeps the image model’s limitations from becoming the video’s limitations.
Where it runs, and how to finish the clip
Nano Banana 2.1 is a Google model. OpenFilmo animates the still it produces.
Nano Banana 2.1 is available through Google: the Gemini app, Google AI Studio, the Gemini API and Flow. Pricing is per image — about $0.0336 for a 1K image, $0.0504 at 2K and $0.0756 at 4K, with batch pricing roughly half — rather than a flat subscription.
To turn a Nano Banana still into a moving clip, bring the image into an image-to-video generator. In OpenFilmo, the image-to-video tool takes the still as the first frame and adds motion, and the result can be upscaled to 4K with no watermark on the Free plan. OpenFilmo does not run Nano Banana itself; it is the animation step after the image is made.
Related guides
Longer reads on the same workflow, from the OpenFilmo blog.
Image to Video AI: Animate Any Photo in Seconds
A complete guide to image to video AI: what makes a good source image, how to write motion prompts, top use cases, and how to fix …
Character Consistency in AI Video: What Actually Works
How to keep characters consistent across AI video shots: reference images, prompt discipline, key frame locking and post-hoc corre…
AI Video Upscaling to 4K: Refiners, Settings and Limits
How AI video upscaling and one-step refiners work, when to upscale versus regenerate at native resolution, and the settings that a…
How to Use Google Veo: A Practical Guide
How to use Google Veo: access routes, prompt approach, working with audio, enterprise considerations, and when the ecosystem advan…
Related workflows
Closely related tools and guides on the same problem.
Frequently asked questions
What is Nano Banana 2.1?
It is Google’s image generation and editing model, released in general availability on 6 October 2026. It produces stills at 1K, 2K and 4K, accepts up to 14 reference images for character and object consistency, and adds mask-based editing and image-search grounding. It generates images and text, and takes video as an input.
Can Nano Banana 2.1 make videos?
No. It is an image model with no video output. To make a clip from its images, animate the still with a separate image-to-video model — for example OpenFilmo’s image-to-video generator. Nano Banana is not built into OpenFilmo; it is the step that produces the frame before animation.
How much does it cost?
Pricing is per image through Google’s API: about $0.0336 for a 1K image, $0.0504 at 2K and $0.0756 at 4K, with batch pricing roughly half those rates. There is no flat subscription, and it is billed by Google, not by OpenFilmo.
What resolutions and ratios does it support?
Output runs at 1K, 2K and 4K, with 1K as the default. The 0.5K size from the previous generation was dropped. Fourteen aspect ratios are supported, including 1:4, 4:1, 1:8 and 8:1 panoramic frames, where tiling artefacts at 2K and 4K were fixed.
Does it keep characters consistent across images?
Yes. Multi-image fusion accepts up to 14 reference images and is documented as holding up to four characters and up to ten objects consistent. That is what makes it useful for storyboards and character sheets, where matching frames matter more than a single pretty still.