This is an AI video generation comparison for
text-to-video 
prompt:
A first-person POV shot using an ultra-wide cinematic lens with slight distortion. The camera moves as if the viewer is slowly walking forward, with their hands visible in front of them, carefully interacting with the environment. The scene takes place on a busy city street during the day, lit by clean natural daylight with sharp shadows and crisp, high-contrast tones. Everything in the environment is completely frozen in time—people are suspended mid-motion, objects hang in the air, and debr...
Log in to see full prompt.
Tested: April 3, 2026
JSON prompt by Pierrick Chevallier x/@CharaspowerAI linked to the original post below
Tested: April 3, 2026
Sent this JSON prompt verbatim, 16:9 widescreen format. Great result, "Matrix" vibes.
Tested: April 3, 2026
OK
Tested: April 3, 2026
Some models interpret this prompt this way with cartoonish people or sometimes uniform creepy grey humanoid figures
Tested: April 3, 2026
Not sure why the quality id so degraded but the physics and prompt understanding, the whole freeze-time implementation is good. Using Sora 2 Fast model.
Tested: April 3, 2026
Don't pick this model for this kind of a prompt. Time-freeze is about the only thing it's done right in here
Tested: April 4, 2026
This model is 'getting there', much stronger than 2.5 but its successor 3.0 does this with superior understanding. This one is unable to overcome uniformity in crowds. But freeze-time effect and POV is already solid.
Is the POV perspective consistent, with hands naturally visible in frame?
Does the environment clearly show a busy city street in daylight?
Are people and objects convincingly frozen mid-action (no subtle drifting)?
Are suspended elements (debris, objects) held perfectly still before release?
Do the hands interact with objects without warping or clipping issues?
Is anatomical motion of the hands/fingers correct during manipulation?
Does the ultra-wide lens produce subtle, consistent edge distortion?
Are lighting and shadows stable and consistent during the freeze?
Are time-freeze particles visible and coherent in the scene?
When time resumes, do all elements snap into motion simultaneously?
Does the release show correct physics (gravity, momentum, collisions)?
Do clothing and hair transition from stillness to natural motion on release?
POV walk through a frozen city where time suddenly snaps back - good perfect for testing AI control over physics, motion continuity, and cinematic time-freeze effects. JSON prompt shared by Pierrick Chevallier x/@CharaspowerAI.
Check out the results from Wan (Online Platform) (Wan 2.7) vs Flova.ai (Seedance 2.0) vs Freepik (Kling 3.0) vs Freepik (PixVerse V5.5) vs Freepik (Sora 2) vs Freepik (Hailuo 2.3) vs Freepik (Kling 2.6) for similar or identical prompts side-by-side.
Split scene cat jumps from fryer to rock show
Holiday chaos bullet time effect
