You know how fast AI video is moving, right? It’s absolutely wild! I just had to see what the latest tools could do, so I decided to throw a T-rex at them. Literally. 🦖
I pitted six of the hottest new models: Kling, Veo, Hailuo, Midjourney, Seedance, and Luma Ray2, against each other in an image-to-video showdown.
But I didn’t make it easy. I gave each one the same image and a complex, 4-part story prompt to see who could really follow directions:
- 1️⃣ Show the T-rex chasing the jeep.
- 2️⃣ The T-rex attempts to bite the jeep.
- 3️⃣ The jeep accelerates and flees.
- 4️⃣ Zoom in on the sad face of the T-rex.
💡 My Big Takeaways
Overall, the results were awesome, but here’s what stood out:
- Longer is better: The models that could create 8-10 second videos, like Kling 2.1, Veo 3, and Hailuo 02, followed the prompt like champs.
- They’ve got personality: It’s cool seeing the models inherit traits from their image-gen families. Hailuo, for instance, has a slick, animation-style movement.
- The surprise hero: Among the shorter 5-second clips, Seedance was the standout! It was the only one that performed all four actions, which is super impressive for a “mini” version.
My personal favorite in this test? Veo 3 just crushed it. But honestly, the progress across the board is a game-changer. A year ago, this kind of action sequence was impossible!
Check out the full Linkedin post to see all the video results for yourself!