AI video comparison: what the tools do and cost

Описание: I keep seeing articles with the headline:
“The best AI video tool of 2025”

These are usually pure clickbait titles - because in my opinion, no tool is “the best”.
Every platform has its strengths, weaknesses and, above all, different purposes.

Nevertheless, comparisons are not completely useless.
Done properly, they provide a good insight into the current state of the art - and that was precisely my aim with this test.

🔍 My setting:
I chose an image from my surreal series for the comparison - with deliberately complex content:
(image and video prompts can be found below)

Difficult prompt realization
Movement & consistency of the figure with mask
Preservation of small details (e.g. squeaking duck)
Physics & dynamics in the video

The results are exciting and sometimes very different. The price-performance ratios are also particularly interesting.

🔸 Midjourney
Good prompt conversion, dynamic, good consistency
Cost: approx. 0.25 (480p)

🔸 Kling AI 2.1 Pro
Somewhat exaggerated dynamics, glitches in the duck
Cost: approx. 0.35 (1080p)

🔸 Wan 2.2
Good realization, a bit silly, overall solid
Cost: approx. 0.20 (720p)

🔸 Hailuo AI 2.0
Best prompt realization, good consistency, fin on the back disturbs
Cost: 0.88 (1080p) or 0.25 (720p)

🔸 Google Veo 3
Good dynamics, good consistency, duck doesn't bounce properly
Cost: approx. 5 (720p)

🔸 Runway Gen-4
Quasi still image.
Cost: approx. 0.45 (720p) (via Freepik)

🔸 Pixverse 4.5
Rating: Good prompt conversion, dancing dynamics, good consistency, lacks sharpness
Cost: approx. 1.15 (720p) (via Freepik)

🔸 Seedance 1.0 Pro
Rating: Prompt realization ok (cough missing), good dynamics and consistency
Cost: 1.10 (1080p) (via Freepik)

🏁 My conclusion:
There is no real “winner”.

But:
📌 Kling AI impresses with its overall look and feel
📌 Wan 2.2 is visually witty and efficient
📌 Hailuo AI offers strong prompt fidelity

👉 Runway Gen-4 falls significantly short in direct comparison.

The costs depend on the selected plan or API usage - the values given here are only a rough guide.

How do you rate such comparisons? And is price or performance more important to you?