Which model gives you the most believable ANIMAL MOTION - the weight, the gait, the muscle?

by @nikolaimehta in c/nature-docs · · Wildlife

Not scenery - specifically a four-legged animal moving with correct biomechanics (a running cat, a walking elephant). Who handles weight and gait best right now?

  • Kling (weight + gait) (15 votes)
  • Hailuo / MiniMax (18 votes)
  • Runway Gen-4 (16 votes)
  • Seedance (17 votes)
  • Veo (14 votes)
  • Sora (18 votes)

98 votes · poll closed

29 points · 7 comments

Comments

  • @rileytanaka_vfx · · 70 points

    Seedance surprised me on birds specifically - wingbeat timing and the way a landing bird backpedals its wings to brake. different models have different animal strengths, worth testing per-creature.

  • @nikolaismith_films3969 · · 48 points

    model-per-creature is real. I've started keeping notes: Kling for big cats, Seedance for birds, Hailuo when I need a weird invented gait. no single winner.

  • @miapark_creates54 · · 98 points

    Kling for quadruped weight, hands down. It understands that an elephant's foot COMPRESSES on contact and the mass settles. Most models float animals like they're on the moon. The weight sells the realism more than the fur detail.

  • @reesesantos_ai · · 90 points

    the moon-gait problem is SO common. a 'majestic lion' that moves like it weighs nothing. weight is everything in wildlife - a real predator's walk has coiled mass. Kling gets closest to that coiled heaviness.

  • @rileymartinlab · · 65 points

    galloping is where they ALL still fall apart. the four-beat gallop gait is biomechanically specific and most models give you a weird two-beat rocking-horse thing. horses remain cursed.

  • @lunasilva_ai4633 · · 60 points

    the rocking-horse gallop haunts me. Muybridge solved this in 1878 and the models still can't do a correct gallop sequence. we've regressed 150 years for one specific gait lol