avatar

For those who haven't heard of it — Happy Horse 1.0 currently sits at #1 on the Artificial Analysis Video Arena leaderboard with an Elo of 1333 (text-to-video, no audio) and 1392 (image-to-video, no audio).

Key highlights:

  • 15B unified Transformer architecture (40 layers, self-attention only, no cross-attention)
  • Jointly generates video + audio in a single pass
  • First/last frame image-to-video control
  • Native support for 6 languages with lip-sync for Mandarin/Cantonese
  • 1080p capable, 5s video in ~38s on H100

We've just added it to lovegen.ai — you can try it right now with free credits. Both text-to-video and image-to-video modes are supported.

→ https://lovegen.ai/happy-horse-1

Login to comment.