I seriously thought Google’s latest video AI update, going from v3.0 to v3.1, was just going to be a minor tweak. It felt like one of those small number changes that don’t mean much. Then I stumbled upon this incredible deep-dive from this AI professional, and my jaw just about hit the floor!
The expert revealed that Veo 3.1 isn’t just a simple quality bump. It’s a massive expansion of creative control, allowing for some truly mind-bending experiments that even outperform other models in surprising ways.
The creator tested out all the new features, and the results are something else. Here’s what stood out:
- 📌 Next-Level Creative Tools: The biggest update is a new “Ingredients to Video” feature. The person who shared it combined an image of himself, a moose hat, and a candy world to generate a video of him dancing through a sugar-coated landscape. He also tested the new “Frames to Video” tool by trying to morph himself into a wolf, and while it wasn’t perfect, the results were hilariously epic. It shows we’re getting way more granular control over our creations.
- ✅ You Can Finally Edit Videos: Veo 3.1 now lets you add new elements after a video is generated. In one test, the expert added an animated person walking into a scene, and it worked perfectly. Adding a spaceship resulted in a static image, so it’s not flawless, but it’s a huge step forward. The creator noted that trying to change an existing object (like a lightsaber into a hockey stick) failed, and removing objects isn’t available yet, but the potential for iteration is huge.
- 💡 The Uncensored Surprise: This is the part that really shocked me. This industry pro decided to test the guardrails by prompting for trademarked characters. Veo had no problem generating videos of “Mickey Mouse high-fives Super Mario” and an anime scene with “Batman and Spongebob.” What’s crazy is that when he ran the same prompts in Sora, it refused, citing content policy violations. It seems like Veo is currently the more flexible option for fanfiction and pop culture mashups.
The expert’s final take is that while Sora might still have a slight edge on pure realism for a single prompt, Veo’s new editing tools and surprising lack of censorship make it an incredibly powerful and fun model to use.
You have to see the full tests for yourself, so check out the original post for all the wild video examples.