HappyHorse 1.0
Released Apr 28, 2026
Video-and-audio generation model from Alibaba's ATH (Alibaba Token Hub) business unit: multi-shot clips up to 15 seconds at 1080p with synchronized audio, including lip-synced dialogue.
- Price per second
- $0.24
- Max clip length
- 15 s
- Max resolution
- 1080p
- Native audio
- Yes
- AA arena Elo
- 1023
- LMArena Elo
- 1427
- Capabilities
- text to video, image to video, video editing
- Licence
- Proprietary
Strengths & weaknesses
Strengths
- Synchronized audio, including lip-synced dialogue and ambient sound
- Text-, image- and subject/reference-to-video, plus video editing
- Character consistency across multi-shot cuts
Weaknesses
- 15-second maximum
- 720P/1080P only on 1.0 (1.1 adds 480P)
- API-only, no open weights found
Developed by Alibaba's ATH (Alibaba Token Hub) business unit, per Alibaba's own post; filed under the existing 'alibaba' company id. releaseDate is the limited-beta launch in Alibaba's 2026-04-28 post (originally on Alizila). Press reports say it first appeared anonymously on the Artificial Analysis Video Arena around 2026-04-07 and that Alibaba confirmed it about 2026-04-10; no primary source verified. A newer HappyHorse 1.1 (happyhorse-1.1-t2v/i2v/r2v) is on Model Studio and is the version Alibaba's docs now use in their examples. It ranks outside the AA top 10, so it is skipped. Model Studio lists 1.0 output as '720P, 1080P. 3-15s. 24 fps', with audio. No Hugging Face repo found, so openWeights false.
Links
API model strings
- alibaba-model-studio
happyhorse-1.0-t2v - alibaba-model-studio
happyhorse-1.0-i2v - alibaba-model-studio
happyhorse-1.0-r2v - alibaba-model-studio
happyhorse-1.0-video-edit