Google’s Cheap, Fast AI Image and Video Play: Nano Banana 2 Lite and Gemini Omni Flash

Google just did the thing it does best: it flooded the zone.

On June 30, it shipped two generative media models on the same day. One makes images almost instantly and for pennies. The other lets you generate and edit video by talking to it. Neither is a demo. Both are live in the API right now.

Here is the part most coverage buried. These two are not meant to be used alone. Chain them and you get an image-to-video assembly line cheap enough to run at scale. That is the actual story.

Let’s break down what launched, what it costs, and how you can start building with it today.

What Google Actually Shipped

Two models, announced together on the Google blog:

Nano Banana 2 Lite (technical name: Gemini 3.1 Flash Lite Image). The fastest, cheapest image model in the Nano Banana family. It returns a text-to-image in about four seconds at $0.034 per 1K-resolution image.
Gemini Omni Flash (gemini-omni-flash-preview). A video generation and conversational editing model, now open to developers for the first time at $0.10 per second of video output.

Both are available in Google AI Studio, the Gemini API, and the Gemini Enterprise Agent Platform. Nano Banana 2 Lite is also rolling into consumer products you already use: AI Mode in Search, the Gemini app, NotebookLM, Google Photos, Stitch, Google Flow, and Google Ads.

If the name rings a bell, it should. This is the same family behind the original Nano Banana, the image editor that quietly went viral last year after topping the LMArena charts under a mystery banana emoji. That model is now the legacy one. Google is telling developers to move on.

Nano Banana 2 Lite: Fast Enough to Change How You Work

The headline number is not the price. It is the four seconds.

Four seconds is fast enough to put image generation inside a live loop. A user types, waits a beat, and sees a result. No spinner, no coffee break. Google built Lite for exactly this: rapid ideation and high-volume pipelines where speed and cost are the whole point.

The trade-off is honest. Lite only outputs at 1K resolution, and Google’s own framing puts it below the full Nano Banana 2 and Nano Banana Pro on raw quality. According to VentureBeat, internal notes peg it at roughly 60 to 70 percent of the general capability of the bigger models, while running much faster and cheaper.

What Google says it did not sacrifice for speed: prompt adherence, character consistency, and legible in-image text. That last one matters. Readable text inside a generated image is usually the first thing to fall apart when a model is tuned for throughput. If Lite holds up there, it is genuinely useful for ad creative, thumbnails, and product shots.

Where Nano Banana 2 Lite fits in the family

Google now runs a clear tier system. Here is how the models stack up, with reported pricing per 1K-resolution image:

Model	Technical name	Best for	Reported price / 1K image
Nano Banana 2 Lite	Gemini 3.1 Flash Lite Image	Speed, volume, tight budgets	$0.034
Nano Banana 2	Gemini 3.1 Flash Image	The everyday workhorse	$0.067
Nano Banana Pro	Gemini 3 Pro Image	Complex, professional jobs	$0.134
Nano Banana (legacy)	Gemini 2.5 Flash Image	Being retired	$0.039

The takeaway is simple. If you were using the original Nano Banana, swap in 2 Lite and you get better quality, faster speeds, and a lower bill in one move. Pricing beyond the Lite figure comes from VentureBeat’s reporting rather than Google’s own post, so treat the exact cents as a guide, not gospel.

Gemini Omni Flash: Editing Video by Talking to It

This is the more surprising launch.

Gemini Omni Flash first appeared at Google I/O 2026 as a preview. Now developers can actually build with it. It generates video and, more interestingly, edits it through plain conversation. You describe a change, it makes the change, and you keep going.

What it does well, in Google’s words:

Conversational editing. Refine a clip using natural language instead of a timeline and keyframes.
Multimodal referencing. Feed it text, images, and video together to keep a scene consistent.
Real-world knowledge. It pulls on Gemini’s understanding of physics, history, and narrative logic to build clips that make sense.
Text and action sync. Connect on-screen graphics to what is happening in the video with a prompt.

The pricing is the clever bit. Omni Flash costs $0.10 per second of video output, the exact same rate as Veo 3.1 Fast. Google draws a real line between the two: Veo is for one-shot, high-quality clips, while Omni Flash is for iterative editing that stitches multiple assets together. Different jobs, same per-second cost.

The Real Move: Chaining the Two Together

Here’s the thing. On their own, a fast image model and a chatty video model are nice. Together, they are a production line.

The flow Google is pushing looks like this:

Generate a reference image with Nano Banana 2 Lite in about four seconds.
Pass that image to Gemini Omni Flash as a reference.
Animate it into a 10-second clip.
Refine the clip through conversation, stacking up to three sequential edits while the model keeps context.

That context-keeping runs on Google’s Interactions API, which holds session history so each edit builds on the last instead of starting fresh.

What this really means: image-to-video creative that used to need a small team and a real budget can now run as an automated pipeline. Product shots that become e-commerce clips. Room photos that turn into interior design walkthroughs. Selfies that drop you at a landmark and then animate the scene. Those are not hypotheticals. They are the three demo apps Google shipped alongside the models.

Try It Yourself in 10 Minutes

You do not need to be an engineer to poke at these. Google AI Studio runs in a browser and needs no setup. Here is the fastest path to a first result.

Open Google AI Studio. Go to aistudio.google.com and sign in with a Google account. It is free to start experimenting in the playground.
Pick Nano Banana 2 Lite. In the model selector, choose the image model (Gemini 3.1 Flash Lite Image). Type a clear, specific prompt. The more detail you give on subject, style, and lighting, the better the result.
Generate and time it. Hit generate and watch how fast it lands. Try a prompt with words in the image, like a poster or a labelled product, to test the in-image text rendering for yourself.
Switch to Gemini Omni Flash. Select the video model (gemini-omni-flash-preview). Upload the image you just made as a reference.
Describe the motion. Tell it what should move. Something like “slow push-in on the subject, soft morning light, gentle camera drift.” Keep the first try simple.
Edit by chatting. Once you have a 10-second clip, ask for a change in plain language. “Make it warmer.” “Add a slow zoom.” Stack up to three edits and watch the context carry over.
Check the guides if you get stuck. Google published a Nano Banana prompting guide and an Omni Flash guide with example prompts.

A useful first project: take one product photo, generate three background variations with Lite, then animate the best one into a short clip with Omni. That single loop tells you more about these models than any benchmark chart.

What This Means If You Make Things for a Living

For marketers and small teams, the cost math is the story. At roughly three cents an image and ten cents a video second, testing 50 ad variations is no longer a budget line you have to defend. You can generate, animate, and A/B test creative at a volume that was priced out of reach a year ago.

For founders building products, the four-second latency is the unlock. It is fast enough to sit inside a consumer feature where a user is actively waiting. That changes what you can put in front of people without them bouncing.

For creators, the conversational editing lowers the skill floor. You do not need to learn a video timeline to make a decent clip move. You describe what you want and refine it like you are texting an editor.

And for the wider generative video market, this is Google doing what it did with the original Nano Banana: pricing a capability low enough that it becomes the default rather than the premium option.

The Honest Limitations

Before you rebuild your whole workflow around this, keep a few things straight.

Omni Flash is a public preview, not a finished product. It has real gaps, and Google says so openly.
Clips max out at 10 seconds. Longer durations are coming, but they are not here yet.
Audio references and scene extension are not supported in the API yet. Plan around that.
Video references up to three seconds are accepted by the API but not processed correctly. The schema takes them; the model ignores them for now.
Character consistency wobbles. Change scenes or pan the camera and the model can lose track of a face or object. Google says it is working on it.
Lite is 1K only. If you need 2K or 4K, you want the full Nano Banana 2 or Pro instead.
The numbers are Google-stated. The speed and quality claims come from Google’s own benchmarks. Test them on your own prompts before you trust them.

One more thing worth knowing: both models tag their output with SynthID watermarking, and you can verify AI content through the Gemini app, Gemini in Chrome, or Search. Good to remember before you pass anything off as camera-shot footage.

The Bottom Line

Read these two releases as one. A fast, cheap image model and a conversational video model, designed to snap together into a single image-to-video pipeline. The individual specs are fine on their own. What they unlock together is high-volume creative production at a price that finally makes sense.

So prototype eagerly. Just ship deliberately, with a human checking what actually goes out the door. Preview models are generous with promise and stingy with polish.

The forward read is the same one that followed the first Nano Banana. When Google makes something this cheap and this fast, it does not stay a novelty for long. It becomes the water everyone else has to swim in.

If you know a marketer still paying a studio to animate product shots one at a time, send them this. Their next quarter’s creative budget just got a lot more interesting.

Sources and further reading

Google AI for Developers, Gemini Omni developer documentation.

Google, Start building with Nano Banana 2 Lite and Gemini Omni Flash (June 30, 2026).

VentureBeat, Google unveils Nano Banana 2 Lite aka Gemini 3.1 Flash-Lite for low-cost, 4-second image generations.

The Decoder, Google launches Nano Banana 2 Lite for fast AI images and Gemini Omni Flash for video via API.

Neowin, Google announces Nano Banana 2 Lite image generation model targeting high-volume workflows.