by Alibaba ATH AI · #1 on Video Arena · Coming soon to imageat
Happy Horse 1.0VIDEO + AUDIO IN ONE PASS
Native 1080p · Joint audio-video generation · 7-language lip-sync.
Ranked #1 with 1362 Elo — 94 points clear of any other model.
Showcase 1
Showcase 2
Showcase 3
Showcase 4
Showcase 5
Showcase 6
Showcase 7
Showcase 8
Showcase 1
Showcase 2
Showcase 3
Showcase 4
Showcase 5
Showcase 6
Showcase 7
Showcase 8
Sample clips — Happy Horse 1.0 by Alibaba ATH AI (coming to imageat)
#1 on Artificial Analysis Video Arena · April 2026 · 94 Elo points ahead of #2
Leaderboard
#1 in blind video testing
Artificial Analysis Video Arena ranks models via blind head-to-head matchups — users vote without knowing which model made each clip. Happy Horse 1.0 leads with 1362 Elo.
Elo ratings from Artificial Analysis Video Arena as of April 2026. Approximate values.
Source: Artificial Analysis Video Arena · April 2026
How it works
Prompt to 1080p video in one pass
No separate audio sync. No upscaling. Just describe what you want.
Write your prompt
Describe your scene, characters, dialogue, and mood. Happy Horse 1.0's unified pipeline handles text-to-video and image-to-video from the same prompt.
Audio generates alongside the video
No separate sync step. Dialogue, ambient sound, and Foley effects are generated in the same pass — in your language of choice.
Download your 1080p clip
Get a production-quality 5–8 second clip at native 1080p, ready for social, marketing, or narrative content.
Choose Your Plan
Get credits delivered to your account every month. Cancel anytime, no questions asked.
Starter Subscription
Get 150 credits every month at a discount
AI IMAGE
- Nano Banana, Nano Banana 2 & Nano Banana Pro
- Turn images into prompts
- AI Image Upscaler (4K)
- Watermark Remover
- AI Influencer Generator
AI VIDEO
- Veo 3.1, Sora 2, Kling 3.0 & Kling 2.6
- Grok Video & Seedance 2.0
- Motion control & end-frame video
- Watermark-free exports
BENEFITS
- Cancel anytime
- No charge for failed generations
- Private generations
- Commercial usage rights
Pro Subscription
Get 350 credits every month for power creators
AI IMAGE
- Nano Banana, Nano Banana 2 & Nano Banana Pro
- Turn images into prompts
- AI Image Upscaler (4K)
- Unlimited Watermark Remover
- AI Influencer Generator
AI VIDEO
- Veo 3.1, Sora 2, Kling 3.0 & Kling 2.6
- Grok Video & Seedance 2.0
- Motion control & end-frame video
- Watermark-free exports
BENEFITS
- Cancel anytime
- No charge for failed generations
- Private generations
- Commercial usage rights
- Priority Support
Premium Subscription
Get 530 credits every month for ultimate production
AI IMAGE
- Nano Banana, Nano Banana 2 & Nano Banana Pro
- Turn images into prompts
- AI Image Upscaler (4K)
- Unlimited Watermark Remover
- AI Influencer Generator
AI VIDEO
- Veo 3.1, Sora 2, Kling 3.0 & Kling 2.6
- Grok Video & Seedance 2.0
- Motion control & end-frame video
- Watermark-free exports
BENEFITS
- Cancel anytime
- No charge for failed generations
- Private generations
- Commercial usage rights
- Priority Support
Business Subscription
Per-seat plan with shared monthly credit pool (1–5 seats)
AI IMAGE
- Nano Banana, Nano Banana 2 & Nano Banana Pro
- Turn images into prompts
- AI Image Upscaler (4K)
- Unlimited Watermark Remover
- AI Influencer Generator
AI VIDEO
- Veo 3.1, Sora 2, Kling 3.0 & Kling 2.6
- Grok Video & Seedance 2.0
- Motion control & end-frame video
- Watermark-free exports
BENEFITS
- Cancel anytime
- No charge for failed generations
- Private generations
- Commercial usage rights
- Priority Support
Trusted by 100,000+ creators worldwide
Why Happy Horse 1.0
Three things no other model does better
Verified by blind testing. Built on a 15B-parameter unified architecture.
Sound that's part of the scene, not added after.
Happy Horse 1.0 generates dialogue, ambient sound, and Foley effects in a single unified pass — no post-processing audio sync. Multilingual support covers English, Mandarin, Cantonese, Japanese, Korean, German, and French.
7 languages65% user preference in blind testing.
Ranked #1 on Artificial Analysis Video Arena with 1362 Elo — 94 points ahead of Seedance 2.0 in second place. Results come from blind head-to-head matchups where users vote without seeing which model generated each clip.
1362 Elo · #1Full resolution, standard aspect ratios.
Happy Horse generates 5–8 second clips at native 1080p in 16:9 and 9:16 — no upscaling, no quality loss. Inference runs in approximately 38 seconds on a single H100 GPU.
1080p · 5–8 secUnder the hood
15B-parameter unified architecture
According to the team, Happy Horse 1.0 uses a 40-layer self-attention Transformer where the middle 32 layers share parameters across text, image, video, and audio tokens in a single unified sequence — no cross-attention branches. The first and last 4 layers handle modality-specific encoding and decoding.
15B
Parameters
40
Layers
32
Shared layers
8 (DMD-2)
Inference steps
Architecture specs are based on team-published information and have not been independently verified. Model weights are not yet publicly available.
What creators say
"The motion in Happy Horse clips feels different — more expressive, less robotic. And the audio coming out in the same generation is a huge workflow win."
Jordan L.
Social media creative
"94 Elo points over Seedance in blind testing. That's not a close race. Can't wait to have this available on imageat."
Aiko T.
Brand video producer
"Multilingual lip-sync in the same generation pass is the feature I didn't know I needed. This changes how we produce localised video content."
Carlos M.
Content agency founder
Happy Horse 1.0 vs other imageat video models
Happy Horse 1.0 — best for: (coming soon)
Overall quality leader by Elo. Joint audio-video generation, multilingual lip-sync (7 languages), native 1080p. Strongest choice for social content, branded video, and any production where audio matters.
Seedance 2.0 — best for:
Director-level camera control and native audio — ranked #2 on the Video Arena. Great for cinematic productions where precise camera movement is the priority.
Kling 3.0 — best for:
Ultra-realistic physics and 4K resolution. The top choice when real-world fidelity and maximum resolution are the goal.
Google Veo 3.1 — best for:
Industry-leading cinematic video with integrated AI audio, up to 4K. Best for high-production-value content with synchronized sound.
Elo rankings from Artificial Analysis Video Arena, April 2026. All models available on imageat — try them now.
Coming soon to imageat
Happy Horse 1.0 is on its way
While you wait, generate AI video today with Kling 3.0, Seedance 2.0, Google Veo, and more — all on imageat with no API keys needed.