Sample
MiniMax
video model

Hailuo 2.3

MiniMax's video model is built around realistic motion and expressive character work — flips, dance, and micro-expression changes read as natural rather than uncanny, across a wide range of visual styles.

700 crText promptReference image

Excels at

  • Strongest human physics/expression fidelity here (flips, dance, micro-expressions)
  • Wide style range: anime, photoreal, cel-shaded, illustrated
  • Consistent character motion across the full clip length

Signal

Use when

  • Character-driven clips needing expressive, natural movement
  • Stylized content (anime, cel-shaded) as well as photoreal
  • Fixed-length clips where consistent motion matters more than long duration

Avoid when

  • Very long or multi-shot sequences (fixed 6s/10s clip lengths)
  • Native multi-reference-image workflows (use Seedance 2.0)

Prompt used for this sample

A paper boat drifting down a rain-slicked city street at dusk, gentle camera follow, cinematic

Also available

MiniMax H3 — The flagship successor — native synced audio, a much larger reference cast, and first-to-last-frame control

Related models

Sample
Kuaishou
video

Kling v3

Fluid, believable human motion for action and dance.

280 cr/secSee details
Sample
ByteDance
video

Seedance 2.0

Multi-shot cinematic sequences with native audio.

1,000 cr/secSee details
Sample
MiniMax
video

MiniMax H3

Native synced audio, multi-reference, and continuation workflows.

325 cr/secSee details