4 Mind-Blowing Things You Can Do With Veo 3

I was skeptical about AI video. Sora impressed everyone but seemed more like a tech demo than a tool. Runway was cool but limited. Then I tried Veo 3 and actually said "holy shit" out loud.

Google's latest video model is genuinely usable. Here are the 4 ways I've been using it.

1. Creating b-roll for content

This is the most practical use case. Need 5 seconds of someone typing at a laptop? A cityscape at sunset? Coffee being poured in slow motion?

Veo 3 generates it in minutes. The quality is good enough for YouTube videos and presentations. It's not replacing professional footage, but for quick projects? Game changer.

I generated a 4-second clip of "morning coffee steam rising in golden light" and it looked like stock footage I'd pay $50 for.

2. Prototyping video ads

Before spending budget on actual production, I now prototype video ad concepts with Veo 3. Generate a rough version of the concept, see if it works, iterate before any real money is spent.

Last week I tested 6 different ad concepts in an afternoon. In the old world, that would've been weeks of pre-production.

3. Social media content

Short clips for Instagram, TikTok, Twitter—anywhere you need quick video. Veo 3 generates 5-8 second clips that are eye-catching without looking like obvious AI slop.

The key is keeping it short. The longer the clip, the more likely something looks wrong. But for social snippets? Perfectly usable.

4. Explaining visual concepts

When I need to show "what this could look like" to a client or team, Veo 3 creates quick visualizations. Instead of describing "imagine a user scrolling through the app," I can generate a rough version.

It's not final product quality, but it communicates ideas better than static mockups.

---

Where Veo 3 still struggles

Let's be honest about the limitations:

Hands and fingers: Still weird sometimes. Better than last year's models, but complex hand movements can look off.

Text in video: Hit or miss. Sometimes it renders text correctly, sometimes it's AI gibberish.

Long clips: Anything over 10 seconds starts having consistency issues. Characters change slightly, physics gets weird.

Specific faces: You can't reliably generate specific people. This is arguably a feature (fewer deepfakes) but limits certain use cases.

The bigger picture

AI video crossed a threshold with Veo 3. It went from "interesting tech demo" to "tool I actually use weekly." The trajectory is clear—in a year or two, this will be standard for content creation.

If you want to try Veo 3 alongside other AI video tools (Runway, Kling, Sora), LazySusan has them all in one place. $2 gets you a week to experiment. Worth it to see where AI video actually is right now.

What's next

I'm most excited about combining AI video with AI audio. Generate visuals with Veo, add voiceover with ElevenLabs, music with Suno—the whole production pipeline is getting automated.

We're not there yet, but we're close. Veo 3 is another step toward anyone being able to produce video content that used to require teams and budgets.