I’ve spent the last few weeks bouncing between four tools that keep showing up in every “best AI video generator” list this year, and I’ve got some thoughts. Google just dropped two new models at the end of June — Nano Banana 2 Lite and Gemini Omni Flash — right into a field where Kling and Veo already had loyal followings. So let’s talk about where each one actually shines, and where I’d still reach for something else.

• Nano Banana 2 Lite: the “just give me something fast” model

I’ll be blunt — Nano Banana 2 Lite isn’t trying to wow you with jaw-dropping detail. It’s trying to get out of your way. Google built this thing for speed and volume, and it delivers: a text-to-image generation lands in about four seconds, and at roughly $0.034 per 1K-resolution image, you can burn through dozens of variations without sweating the bill.

Where this actually matters is in the “I need fifty versions of this ad creative by 3pm” moments. If you’re A/B testing product shots, spinning up storyboard frames, or just want to iterate on an idea without waiting around, Nano Banana 2 Lite is genuinely great at that job. It’s replaced the original Nano Banana model, and honestly, the upgrade in prompt-following and character consistency is noticeable even at this “lite” tier.

What it’s not for: anything that needs to be your hero image. If you need pixel-perfect detail or complex reasoning about a scene, step up to the standard Nano Banana 2 or Nano Banana Pro instead. Lite is a sketchpad, not a canvas.

• Gemini Omni Flash: “video editing that finally talks back“

This is the one I find more interesting, honestly. Gemini Omni Flash handles video generation and — this is the fun part — conversational editing. Instead of dragging clips around a timeline, you just tell it what you want changed: make the lighting warmer, add more motion to the camera, swap the character’s jacket. It listens, and it does a surprisingly reasonable job of keeping the rest of the scene intact while it makes the change.

You can chain it with Nano Banana 2 Lite too — generate a still image cheaply, then hand it to Omni Flash to bring it to life. That pipeline is where I think Google’s actually onto something, especially for marketing and e-commerce teams who need a lot of visual content fast.

That said, I want to manage expectations here: this is public preview software, not a finished product. Clips currently cap at 10 seconds, there’s no audio-reference upload support yet, and character consistency can wobble when the camera cuts to a new angle. I wouldn’t wire this into a client-facing pipeline just yet — but for prototyping and quick social content, it’s already useful, and priced at $0.10 per second, it matches Veo 3.1 Fast almost exactly.

• Kling: “still the one to beat for serious video work“

If you want the tool that feels the most like an actual filmmaking instrument, Kling is it. Kling 3.0 introduced what Kuaishou calls an “AI Director” mode — it can generate up to six connected shots in a single pass instead of forcing you to stitch separate clips together, and it holds character identity, clothing, and voice steady across all of them. That’s a real workflow difference, not just a spec bump.

The physics also hold up better than most competitors. Objects don’t float, characters don’t warp when they walk behind something, and native audio sync means you’re not bolting sound on afterward. Clips go up to 15 seconds, and the text rendering inside scenes — signage, captions, logos — actually stays legible, which is rarer than it should be in this category.

The catch: customer support and billing have a rough reputation, with intro pricing that can jump at renewal, and as a Chinese-regulated platform, there’s real content moderation on politically sensitive material. If you’re doing narrative or commercial video work and can live with those trade-offs, Kling is hard to beat right now.

• Veo 3: “the cinematic specialist“

Veo earns its reputation on one thing above all — it makes natural motion look natural. Wind through leaves, water movement, atmospheric lighting — Veo 3.1 nails the stuff that trips up other models. It was also one of the first mainstream tools to get native audio generation right, syncing dialogue, ambient sound, and music to the visuals convincingly.

Where it falls short of Kling is multi-shot storytelling. Veo can extend a scene, but it’s built to deliver one polished, cinematic take rather than a sequence of connected shots. Think hero ad clip or establishing shot, not a six-scene story. It’s also on the pricier end and comes in a confusing tier lineup (flagship, Fast, Light), so know which version you’re actually paying for before you commit.

So which one should you actually use?

Honestly — probably more than one. Nano Banana 2 Lite for fast, cheap image drafts. Gemini Omni Flash if you want to iterate on video through conversation rather than a timeline, keeping in mind it’s still preview-grade. Kling if you need a genuinely coherent multi-shot narrative with strong character consistency. Veo 3 if a single, gorgeous, physically-convincing shot is the whole assignment.

None of these tools do everything perfectly, and that’s fine — the smart move in 2026 is picking the right one per shot, not per subscription.