This is an AI video generation comparison for image-to-video 
prompt:
Quickly push in on the woman's face while she blinks as clouds move behind her. Dolly zoom in into one of her eyes, showing the intricate design of her blue iris with black pupil in the center. Surreal macro shot of the iris and pupil becomes animated as tiny black birds emerge from the black pupil and continue flying around in the iris flapping their wings. Artistic, fashion, surreal, dynamic
Tested: June 11, 2025
Beautifully realistic
Tested: June 11, 2025
Tested: June 11, 2025
Using 'Eyes In' effect
Tested: June 11, 2025
None of the 5 generations produced tiny birds in eye, only large ones.
Tested: June 12, 2025
Tested: June 12, 2025
Turbo model again does better in prompt understanding and following than 2.2 which generated tons of 3rd eye on the nose nightmare fuel. Turbo repeatedly understood this prompt and its variations as intended.
Tested: June 19, 2025
All 4 results showed good understanding of the prompt. Dislike the skin texture and added freckles, left and righ eye don't match re makeup and eyelash volume. Birds start from outside the pupil instead of emerging from there,
Does the shot begin with a fast push-in toward the woman's face?
Does she visibly blink during the push-in?
Does the camera perform a dolly zoom into one of her eyes?
Does the shot transition into a surreal macro close-up of the eye?
Do tiny black birds emerge from the pupil, flapping their wings and moving around?
Is the overall visual tone artistic and surreal (not realistic or clinical)?
This is a tricky prompt, especially the tiny birds part. Oftentimes AI will spawn a third eye.
Check out the results from Wan (Online Platform) (Wan 2.1) vs Dreamina AI (Seedance 1.0) vs Higgsfield AI (Higgsfield-v1 Lite) vs Kling AI (Kling 2.1) vs Hailuo AI x Minimax (Hailuo I2V-01) vs Pika (Pika Turbo) vs Midjourney (Midjourney V1) vs Veo (Veo 2) for similar or identical prompts side-by-side.
