What did we choose for the video?

[md.a.z.i.z.ulha.kim4]

We compared three neural networks — Runway , Kling , and Minimax . We used different versions of each for testing, generating all possible variants. These neural networks, in our opinion, are among the top in their field.

Comparison of video neurons

Speaking of close-ups of people, here’s our example with a boy who was supposed to just move forward. Runway left the image unchanged — no lighting, no blurring. Kling significantly degraded the quality: the eyes became barely noticeable, and the picture itself was blurry. Minimax basically “decided” that the boy should leave the frame, meaning they got a non-working version.

Comparison of video neurons

Regarding the animal animation, the results were also different. The cat brazil telegram screening had to meow and then either close its mouth or continue the emotion, for example, hiss. Runway did it well, but added an extra tooth that had to be removed during editing. Kling added a tongue to the cat, which made it look too “human”. And Minimax generally made the cat so that its behavior became unacceptable. In the end, we used the generation from Runway.

Comparison of video neurons

For animation of complex actions, such as a girl throwing a bouquet, Runway again showed the best results.