Character Reference Showdown: Hailuo vs Pika vs Luma Dream Machine vs PixVerse

แชร์
ฝัง
  • เผยแพร่เมื่อ 28 ม.ค. 2025

ความคิดเห็น • 18

  • @TashaCaufield
    @TashaCaufield 15 วันที่ผ่านมา +1

    Fantastic comparisons! Thanks!! 🤗

  • @voodoochild420ai
    @voodoochild420ai 15 วันที่ผ่านมา +1

    cool comparison video with some great examples :)

  • @alexbon_com
    @alexbon_com 15 วันที่ผ่านมา +1

    Thanks for the review! Really helpful ❤

    • @aivideoschool
      @aivideoschool  15 วันที่ผ่านมา +1

      Glad it was helpful!

  • @eudimaparcas
    @eudimaparcas 12 วันที่ผ่านมา +1

    muito obrigado,valeu!!!!

  • @artvsmachine
    @artvsmachine 15 วันที่ผ่านมา +3

    The winner is Vidu. Lets you use 3 reference images. But that's if you like their style. Kling also has an option but it requires a ton of training on videos and is limited to something you have videos of, not generated imagery.

    • @aivideoschool
      @aivideoschool  15 วันที่ผ่านมา +1

      Cool thanks for the tip on Vidu. Sounds like Pika's Ingredients which I love so I'll check it out. I'm hopefully going to do a separate one on Kling because, like you said, they require video reference rather than a single image. Thanks!

  • @Lhukka
    @Lhukka 12 วันที่ผ่านมา +2

    essa plataforma cria vídeos e imagens

  • @lionelson7098
    @lionelson7098 15 วันที่ผ่านมา +3

    Unfortunately, I had a poor experience with this feature of Hailuo, as it frequently produced bizarre artifacts, distortions, and severe disproportions, particularly with the characters' heads. The speed was inconsistent, and the results were far from what was advertised in the promotional material

    • @aivideoschool
      @aivideoschool  14 วันที่ผ่านมา

      Just out of curiosity, were the characters you were using realistic? I ask because I almost tried testing a non-human or illustrated character.

    • @lionelson7098
      @lionelson7098 14 วันที่ผ่านมา

      @@aivideoschool I used human characters, like a child in ninja outfits. Obviously, the AI struggles significantly with dynamic and action-packed videos, especially fight scenes, but I found its issues to be quite exaggerated

    • @aivideoschool
      @aivideoschool  14 วันที่ผ่านมา

      Got it. I agree, it should only generate the action to a level where the face is still present/similar.

  • @kjabasini
    @kjabasini 6 วันที่ผ่านมา

    Amazing thank you , what about multi photos training 😮

    • @aivideoschool
      @aivideoschool  4 วันที่ผ่านมา

      I think only Pika does two characters 04:00 but Kling just released their own version called Elements. You can also do two characters in Kling.

  • @Devin_Kevin
    @Devin_Kevin 14 วันที่ผ่านมา +1

    but they didnt have that kick of emotions in voice and expression and speed of back ground and the focus point is not matching but still i will give 72 its out os 100

    • @Devin_Kevin
      @Devin_Kevin 14 วันที่ผ่านมา

      and salute to you sir 🫡 2 years doing ai only is no shot of small deal understanding ai mechanics especially

    • @aivideoschool
      @aivideoschool  11 วันที่ผ่านมา

      I think the best way to get emotion is using ElevenLabs voice to voice, combined with Runway Act One. Anything purely generated is a hard sell, but if there's a human performance driving it, it's much better.