The Best AI Video Generator for Faking Will Smith Crowds

In partnership with

Looking for unbiased, fact-based news? Join 1440 today.

Join over 4 million Americans who start their day with 1440 – your daily digest for unbiased, fact-centric news. From politics to sports, we cover it all by analyzing over 100 sources. Our concise, 5-minute read lands in your inbox each morning at no cost. Experience news without the noise; let 1440 help you make up your own mind. Sign up now and invite your friends and family to be part of the informed.

Subscribe to 1440 today.

The Best AI Video Generator for Faking Will Smith Crowds

Will Smith’s post-slap fall from grace continued earlier this month when the singer was accused of using AI to fake crowd scenes in a video posted to his Instagram.

I have my other theories about what actually happened, which I’ve shared in an interview with a major news source—hopefully I’ll have a story to share soon!

But in the meantime, I wondered what it would take to create convincing fake Will Smith crowd scenes. In general, it seemed like a great test case for the fairly challenging task of image-to-video generation with AI.

So, I took a real Will Smith crowd photo, and tried using it to create fake crowd videos with three major AI video generators—Runway, OpenAI’s Sora, and Google’s Veo3.

There was a clear winner—and a clear (and unexpected) loser that was so hilarious I could barely narrate my YouTube video without losing it. Here’s the video (or watch on YouTube).

Yes, this is a silly example! But it shows the state of AI video generation today, and makes for a great real-world test of the three models.

Why? Crowd scenes are one of the hardest kinds of images/videos for AI generators to handle. They involve multiple people, faces, movement, and often text (on signs, as in the Will Smith videos).

Creating an AI video of a crowd is thus a great way to battle-test video generators. And again, some did great—others were so bad it was painful.

Why Care About GEO?

When I talk about Generative Engine Optimization, I almost always get at least one question like “Our Chatbot traffic is so tiny, why should I put money into ranking in chatbots?”

If you’ve ever tried to convince your marketing department to give you budget for GEO work, you might have heard this, too!

The naysayers aren’t entirely wrong—referral traffic from LLMs is still tiny, with Ahrefs reporting that only .17% of a typical site’s traffic comes from ChatGPT, Perplexity etc.

But traffic doesn’t tell the whole story. You still need to care about GEO. Here are 3 reasons why (watch on YouTube).