Six AI video models, one product shot

We gave six video models the same start frame and the same prompt, asked for 5 seconds of vertical video, and changed nothing else. Here’s what each one made and what it cost in credits.

The input

PromptSlow cinematic push-in toward the green FERNHILL cold brew can on the mossy rock. Morning mist drifts across the stream, sunlight flickers through the pines, a drop of condensation runs down the can. The can and its label stay exactly the same. Sound: a gentle forest stream and a few birds, no music, no voice.

9:16, 5 seconds, sound on. The frame comes from our ad campaign made in Claude.

The six results

• Gemini Omni Flash

  720p · 85 credits

  A gentle push-in. The label stays sharp throughout.

• MiniMax H3 Max Fast

  768p · 40 credits

  A clear push-in, and the drop of condensation runs down the can as asked. The cheapest by far.

• Seedance 2.0 Mini

  720p · 155 credits

  Very little camera movement; close to a still image with a moving stream.

• Wan 3.0

  720p · 100 credits

  A gentle push-in. The quietest soundtrack of the six.

• Kling 4.0 Omni

  720p · 157 credits

  A push-in with the condensation drop running down the can. The loudest, most present stream sound.

• Seedance 2.5

  480p · 221 credits

  The strongest move: it ends on a close-up of the label, with the drop visible. Rendered at 480p, so it’s the softest picture.

Cost for the same 5 seconds

The cheapest, MiniMax H3 Max Fast, cost 40 credits; the most expensive cost 221. All six kept the can and its label recognisable.

How to use this

• Draft on MiniMax H3 Max Fast: at 40 credits you can try several prompts for the price of one premium clip.
• Use a premium model for the final hero shot once the motion is right.
• Seedance 2.5 ran at 480p here to keep the cost down; at 720p it’s about 473 credits for 5 seconds.