This is an AI video generation comparison for image-to-video prompt:
Medium shot from behind — a man facing a mirror tries on a new outfit. He studies his reflection with a thoughtful gaze, turns slightly to the left, keeping his eyes fixed on the mirror examining how clothes look from the side as he adjusts his purple sweatshirt slightly .
Generated: May 27, 2025
Veo 2. Clearly try-on going on. Very good.
Generated: May 26, 2025
This Pika Turbo's generation is the winner amongst 10 outputs I've generated using Turbo and 2.2 model. As often the case, Turbo often gets things better than 2.2, they should really bump up its resolution and sell it as a superior model. Overall, Pika struggled with this concept.
Generated: May 26, 2025
Runway Gen 4 model.
Generated: May 26, 2025
Kling 1.6 generation is quite good and well synced.
Generated: May 26, 2025
Dreamina's Seaweed Alpha model can't get the synced mirror image quite right, and for some reason throws in a third person or an object obstructing the view form time to time.
Generated: May 26, 2025
FramePack definitely gets the idea of mirror reflection videos, however here it decides to adjust the hair not clothes.
Generated: May 27, 2025
Wan2.1 Plus here ticks all the boxes.
Generated: May 28, 2025
Hailuo's base image to video model I2V-01 with optimisation off.
Mirror reflections are tricky concept for AI to generate, but things seem to be getting better. Let's test with this promt and see:
Does the reflection accurately match the man's pose and outfit?
Does he appear to be examining himself with a thoughtful expression?
Is the man’s body at any point turned slightly to the left?
Are his eyes fixed on the mirror?
Check out the results from Veo, Pika, Runway, Kling AI, Dreamina AI, FramePack, Wan (Online Platform), Hailuo AI x Minimax for similar or identical prompts side-by-side. Note generation dates and comments about models used.