Gemini Omni Flash & Nano Banana 2 Lite

AI video generation has had a one-way problem. You prompt, you generate, you watch. If the output is 90% right, you generate again and hope. The footage you already have, whether AI-generated or shot on a set, has been off-limits: editing it meant leaving the AI workflow entirely.

Dubbed “Nano Bana for Video” by internet AI aficionados, Gemini Omni Flash changes all that: it edits existing footage directly, guided by prompts and image references. Google's newest video model is live in Pencil, and alongside text-to-video and image-to-video generation. It arrives alongside Google’s new Nano Banana 2 Lite, the fastest image model in the Nano Banana family. Both models were officially unveiled at Google’s recent IO 2026 event.

Generate Video. Then, for the First Time, Edit It.

• ‍Text-to-Video. Describe the shot, get the footage.‍
• Image-to-Video. Use an image as a start frame, an end frame, or a reference, one per generation. Turn a product shot into motion, or steer a scene toward a visual you already trust.‍
• Video editing and guidance. This is what makes Gemini Omni Flash special. Combine a video, a prompt, and optionally a reference image to change footage directly. Swap a subject in the video for a character from the image. Keep the shot, change the detail. The 90%-right generation stops being a re-roll and becomes an edit.

Concept at Speed with Nano Banana 2 Lite

Nano Banana 2 Lite is the new entry tier in Google's Nano Banana family, built on Gemini 3.1 Flash Lite and sitting below Nano Banana 2 and Pro. It generates images in a few seconds, much faster than Nano Banana 2.

That speed changes how you use it. Exploring ten visual directions costs seconds, not sessions. Concept fast and wide with Lite, then hand the winning direction to Nano Banana 2 or Pro for the final, highest-fidelity render.

Where You Can Use Them

• Editor: select Gemini Omni Flash in the Video panel.
• Workflows: select it in the video generation node, so editing and generation both scale across a full production run.
• Agents: select it in Video generation, or any agent with video tools added.

Nano Banana 2 Lite works everywhere images are generated: Editor, Canvas, Workflows, Feeds, and Agents.

How to Turn Them On

If your workspace already has a Google video model enabled, Gemini Omni Flash is switched on automatically. The same goes for Nano Banana 2 Lite in any workspace with a Google image model enabled. Otherwise, an admin enables them under AI Governance → Settings → Video Generation or Image Generation.

Details and Current Limits

Gemini Omni Flash is in preview from Google, so quality and behaviour may change. Video references run up to 10 seconds, one at a time. At times its outputs misinterpret or seemingly ignore a reference. If that happens, try tweaking your prompt and retrying the generation. Video uploads max out at 70 MB. Voice editing and uploaded audio references aren't supported yet.

Why This Matters

For enterprise teams, video has been the most expensive asset to iterate. A change request meant a new generation, a new render, a new review cycle. Editing existing footage with a prompt collapses that loop: keep what works, change what doesn't, stay inside the same workflow, governance, and brand controls as the rest of your creative. And with Nano Banana 2 Lite alongside it, the heavyweight models are saved for the final render, not burned on exploration.

Because it's Pencil, both arrive alongside every other leading model, not instead of one. The Open Garden means you can choose the right tool for the job, as soon as it exists. Workflows and setups that rely on older models stay intact, unaffected unless you choose to swap models out.

‍

Gemini Omni Flash and Nano Banana 2 Lite are live in Pencil now.

Try them today: https://pro.trypencil.com/login