Can AI Video Keep Up with Human Movement? A Futureproof Field Test

In each test, we’ll use the same source image and creative prompt across multiple platforms, focusing on real-world use cases like branded content, product demos, and storytelling.

Test 1: Can AI keep a character running smoothly?

We started with a seemingly simple challenge: motion consistency.

The prompt:

A fit man in a blue sweatshirt and gray shorts jogs along the edge of a quiet beach, with soft waves rolling in behind him and footprints forming in the wet sand. The camera tracks him from the side at mid-speed, capturing the energy and fluidity of his stride in the early morning light, with dynamic camera angles and speed-ramped motion.

The goal: See how well each tool could maintain smooth, natural character movement—without glitches, distortion, or broken frames.

Here’s what we found.

🟢Veo 3

Smooth, fluid, and natural. Footsteps and waves were rendered cleanly. There was a subtle glitch in the arm movement at the very beginning, but it corrected quickly. Bonus points for ambient sound, subtle zoom-outs, and voiceover narration.

🟡Kling

Great sprinting motion, and even added realistic breathing. But after a few seconds the limbs began to deform. The environment looked solid. Ocean sounds were good, though the footstep audio felt too hard for sand.

🔴Runway

Jogging looked jerky and unnatural. Arm movement glitched, hands were deformed. It added footprints but skipped the waves. No audio.

🔴Midjourney

Very early days for their video tool. Motion was glitchy and oddly timed, and the jogger ended up sprinting directly into the ocean. No audio included.

🟡Higgsfield

Sand and waves looked realistic, but the running form was awkward and unnatural. The overall pacing felt like slow motion. No sound.

🟡Luma

Kicked up sand like the reference photo, but the runner moved like it was solid ground. Arm glitches showed up early, and the surface looked flat and artificial. No sound included.

Aspect Ratios Are a Production Constraint

Only Veo 3 and Kling produced videos in a standard 16:9 format. The rest defaulted to square or non-standard ratios, limiting their usefulness for marketing-ready content. For brands planning to repurpose video across platforms, format flexibility is as important as motion quality.

What We Learned

Motion may seem like a basic feature, but it’s one of the hardest things for AI video tools to get right.

In real content production, people notice when movement feels off. Whether it’s a product demo, a brand spot, or a lifestyle scene, jittery limbs, glitched frames, or unnatural pacing can instantly break immersion.

In this test, Veo 3 clearly led the pack, delivering the most cinematic and believable result. Kling showed real promise but couldn’t stay stable. Runway and Midjourney highlighted a bigger issue: there's still a big gap between generating a video and generating one you can actually use.

This was just our first trial, focused on the fundamentals of motion. But it raised a bigger question we’ll continue to explore in future tests:

Which AI video tools are actually ready for production-level creative?