HappyHorse ModelThe weights haven't dropped yet.
Get notified when HappyHorse-1.0 releases weights, API access, or any updates.
No spam. Unsubscribe anytime.
Happy
Horse#1 Open-Source AI Video Generation Model
The 15B-parameter open-source AI video generation model that topped every leaderboard — text-to-video, image-to-video, and joint audio in one unified transformer.
HappyHorse-1.0 generated video
SEE IT IN ACTION
AI Video Generation Samples from HappyHorse-1.0
Real outputs demonstrating text-to-video generation quality at 1080p.
ARCHITECTURE
What Makes HappyHorse Different
A unified multimodal transformer purpose-built for joint video and audio generation — no separate models, no post-processing.
Unified Transformer Architecture
40-layer single-stream transformer with 4 modality-specific layers on each end and 32 shared layers. Text, video, and audio tokens processed in one unified sequence — no cross-attention.
Joint Video + Audio Generation
Generates synchronized dialogue, ambient sound, and Foley alongside video frames in a single forward pass. No post-production dubbing pipeline required.
8-Step DMD-2 Distillation
Reduces denoising from 50+ steps to just 8 without classifier-free guidance, accelerated by the in-house MagiCompiler runtime for real-time inference.
Multilingual Lip-Sync
Native support for English, Mandarin, Cantonese, Japanese, Korean, German, and French with industry-leading low Word Error Rate for digital human content.
1080p Cinematic Output
5–8 second clips at full 1080p in standard aspect ratios (16:9, 9:16) — suitable for social media, advertising, and cinematic production.
Open Source & Self-Hostable
Base model, distilled model, super-resolution module, and inference code released with commercial-use permission. Deploy on your own H100 or A100 GPU infrastructure.
AI VIDEO MODEL BENCHMARK
Artificial Analysis Video Arena Rankings
Blind-tested Elo ratings from thousands of human-evaluated comparisons. HappyHorse vs Seedance, Kling, SkyReels, and PixVerse.
Data from Artificial Analysis Video Arena · April 2026
AI VIDEO MODEL COMPARISON 2026
HappyHorse vs Every Major Video Model
Comprehensive comparison of open-source and commercial AI video generation models — parameters, capabilities, licensing, and availability.
| Model | Developer | Params | T2V Elo | Modes | License | API Price |
|---|---|---|---|---|---|---|
| HappyHorse-1.0 | Unknown | 15B | 1332 | T2V, I2V, Audio | Open Source | Coming soon |
| Seedance 2.0 | ByteDance | — | 1273 | T2V, I2V, Audio | Proprietary | No API |
| SkyReels V4 | Skywork AI | — | 1245 | T2V, I2V | Proprietary | $7.20/min |
| PixVerse V6 | PixVerse | — | 1241 | T2V, I2V | Proprietary | $5.40/min |
| Kling 3.0 | KlingAI | — | 1241 | T2V, I2V | Proprietary | $13.44/min |
| Veo 3.1 | — | — | T2V, I2V | Proprietary | $9.00/min | |
| Sora 2 Pro | OpenAI | — | — | T2V, I2V | Proprietary | $24.00/min |
| WAN 2.5 | Alibaba | 14B | — | T2V, I2V | Apache 2.0 | Self-host |
| LTX 2.3 | Lightricks | 22B | — | T2V, I2V, Audio | Apache 2.0 | Self-host |
| Hailuo 2.3 | MiniMax | — | — | T2V, I2V | Proprietary | $4.80/min |
Data compiled from Artificial Analysis Video Arena, model documentation, and API provider pricing pages. Updated April 2026.