The HappyHorse API on Atlas Cloud connects your application to Alibaba's HappyHorse 1.0 and 1.1 video generation models. Produce clips running anywhere from 3 to 15 seconds at 720p or 1080p, steer results with up to nine reference images, and pick the aspect ratio that fits your product. Atlas Cloud adds Day-0 model availability, reliable uptime, and a simple asynchronous REST integration. Start building today.
Atlas Cloud provides you with the latest industry-leading creative models.
Lowest cost
| Modality | Description |
|---|---|
| HappyHorse-1.1 Text-to-Video API | Generates 1080p video with synchronized audio from a text prompt in a single pass. Version 1.1 brings improved motion expressiveness, stronger prompt adherence, and smarter scene planning for complex multi-shot narratives. |
| HappyHorse-1.1 Image-to-Video API | Animates a static image into fluid video while preserving subject identity and visual consistency. Version 1.1 delivers more physically grounded motion and better frame-to-frame continuity across dynamic compositions. |
| HappyHorse-1.1 Reference-to-Video API | Guides video generation using up to multiple reference images for precise control over style, character identity, and brand elements. Version 1.1's upgraded multi-reference fusion keeps product details and visual fidelity stable across the full clip. |
| HappyHorse-1.0 T2V API (Text To Video) | Transforms detailed text prompts into cinematic video sequences with claimed synchronized audio generation. Leverages unified Transformer architecture for joint video-audio synthesis. |
| HappyHorse-1.0 I2V API (Image To Video) | Animates static images with fluid motion while maintaining visual consistency. Processes reference image latents jointly with text and audio tokens in unified sequence. |
| HappyHorse-1.0 T2V+Audio API (Text to Video with Audio) | Generates complete audio-visual content from text alone — dialogue, environmental sounds, and Foley effects through unified token denoising. |
| HappyHorse-1.0 I2V+Audio API (Image to Video with Audio) | Transforms still images into animated scenes with synchronized soundscapes — cinematic audio accompaniment generated in single forward pass. |
From Arena-topping cinematic quality to native multilingual sound, the HappyHorse API combines unified multimodal generation, lifelike motion, image animation, and an asynchronous workflow behind one pay-as-you-go endpoint.
HappyHorse holds the top spot on the Artificial Analysis Video Arena for both text-to-video and image-to-video, judged by blind human preference. That ranking translates into footage with convincing detail, lighting, and cinematic composition.
Native lip sync spans Mandarin, Cantonese, English, Japanese, Korean, German, and French, with dialogue and Foley generated alongside the video. Localized campaigns no longer need a separate dubbing or sound design pass.
Text, image, video, and audio tokens flow through one unified Transformer sequence instead of a chained multi-stage pipeline. Your integration stays simple because every generation mode shares the same underlying architecture.
When subjects run, collide, or interact, the physics stays believable, a strength behind HappyHorse leading blind image-to-video comparisons on the Video Arena. Sports clips, chase scenes, and dance sequences keep their weight on screen.
Feed one photograph to the image-to-video endpoint and watch it become fluid 1080p motion while subject identity stays intact. Product shots, portraits, and concept art all animate without losing their look.
An asynchronous task workflow keeps generation moving: submit a prompt, poll for status, and retrieve the finished clip without blocking your application. Rapid iteration cycles fit tight campaign and prototyping deadlines.
The same prompt, generated by HappyHorse and other leading video models: cinematic short film and high-end commercial film
Create a 10-second cinematic short film about a street musician and a child meeting in a sunny city square. Scene 1, 0-2s: Wide establishing shot of a bright European-style city square in the late afternoon. A street musician sits near a fountain, playing an acoustic guitar. Warm sunlight, soft shadows, pigeons moving in the background, natural city ambience. Scene 2, 2-5s: Medium tracking shot as a young child walks toward the musician while holding a small red balloon. The musician continues playing guitar. The child’s steps, the guitar rhythm, and the ambient crowd sound should feel synchronized and natural. Scene 3, 5-8s: Close two-shot: the musician smiles and slightly changes the melody as the child stops beside him. The child gently taps the balloon in rhythm with the music. Keep both characters’ faces, clothing, body proportions, and spatial positions consistent. Scene 4, 8-10s: Final cinematic pullback shot: the musician keeps playing, the child laughs, pigeons fly across the square, and the fountain sparkles in the sunlight. The music and city ambience fade naturally. Requirements: - Maintain consistent identity, clothing, and proportions for both characters across all shots - Keep the red balloon, guitar, fountain, and city square layout consistent - Use readable cinematic camera control: wide shot, tracking shot, close two-shot, final pullback - Synchronize audio with action: acoustic guitar, footsteps, child laughter, pigeons, fountain, city ambience - Natural human motion, stable hands, realistic guitar-playing movement - No warped faces, no changing character outfits, no disappearing balloon or guitar - Warm cinematic realism, polished short-film look, 1080p
HappyHorse 1.1
Kling V3.0
Pixverse V6
Create a 10-second cinematic lifestyle commercial for a smart home coffee machine. Scene 1, 0-2s: Wide shot of a bright modern kitchen in the morning. A smart coffee machine sits on a clean countertop beside a white ceramic cup. Soft sunlight enters through the window, warm and natural. Scene 2, 2-5s: Medium shot of a young professional entering the kitchen and tapping the touchscreen on the coffee machine. The screen lights up smoothly. Add a soft button tap sound, quiet machine startup sound, and subtle morning ambience. Scene 3, 5-8s: Close-up of coffee pouring into the cup. Steam rises naturally, the crema forms on the surface, and the machine remains the exact same design, size, color, touchscreen position, and material. The pouring sound should match the liquid motion. Scene 4, 8-10s: Final hero shot: the person lifts the cup and smiles while the coffee machine stays sharp in the background. Camera slowly pulls back, clean premium home appliance commercial look. Requirements: - Maintain the exact same coffee machine design across all shots - Keep the kitchen layout, countertop, cup, lighting, and product scale consistent - Use readable cinematic camera control: wide shot, medium interaction shot, close-up, final pullback - Synchronize audio with action: button tap, machine startup, coffee pouring, soft morning ambience - Natural hand movement, realistic liquid physics, stable product geometry - No warped hands, no changing touchscreen position, no inconsistent product details - Bright cinematic realism, high-end lifestyle commercial style, 1080p
HappyHorse 1.1
Kling V3.0
Pixverse V6
From native audio social clips to multilingual campaigns, reference guided characters, and text driven edits, the HappyHorse API puts every generation mode behind one pay-as-you-go Atlas Cloud endpoint.
Dialogue, ambient sound, and Foley effects arrive in the same pass as your 3 to 15 second clips. Creators ship vertical content for TikTok, Reels, and Shorts without a separate dubbing pipeline.
If a product photo is all you have, image to video animation adds lifelike motion while preserving the original frame. Online sellers turn static catalogs into demo clips ready for listings and ads.
Native lip sync spans seven languages, from Mandarin and English to Japanese, Korean, German, and French. Global marketing teams turn one campaign into localized ads for every market without hiring voice talent.
Up to nine reference images steer reference to video generation, locking character identity, style, and brand elements in place. Episodic storytellers and virtual influencer studios keep faces consistent from scene to scene.
Describe a change in plain text and the video edit endpoint applies it directly to existing footage. Post production teams restyle scenes, swap elements, and repair shots without scheduling a reshoot.
Need to see a scene before the crew shoots it? Arena leading visual quality at 1080p turns scripts into previsualization footage that film teams and agencies use to align on framing and pacing.
| Model | Input Types | Output Duration | Resolution | Audio Generation |
|---|---|---|---|---|
| HappyHorse-1.0 | Text, Image | 5–8s | 1024×1024 | √ |
| Seedance 2.0 | Text, Image | 4~15s | 1024×1024 | √ |
| Kling 3.0 | Text, Image | 3~15s | 256P~4K | √ |
| Wan-2.6 | Text, Image | 5s;10s;15s | 1080P, 720P | √ |
Get started in minutes — follow these simple steps to integrate and deploy models through Atlas Cloud's platform.
Sign up at atlascloud.ai and complete verification. New users receive free credits to explore the platform and test models.
Combining the advanced Happy Horse models with Atlas Cloud's GPU-accelerated platform provides unmatched performance, scalability, and developer experience.
Low Latency:
GPU-optimized inference for real-time reasoning.
Unified API:
Run Happy Horse, GPT, Gemini, and DeepSeek with one integration.
Transparent Pricing:
Predictable per-token billing with serverless options.
Developer Experience:
SDKs, analytics, fine-tuning tools, and templates.
Reliability:
99.99% uptime, RBAC, and compliance-ready logging.
Security & Compliance:
SOC 2 Type II, HIPAA alignment, data sovereignty in US.
The HappyHorse API gives developers programmatic access to HappyHorse, Alibaba's video generation model family that debuted at #1 on the Artificial Analysis Video Arena for both text-to-video and image-to-video. Built on a unified multimodal Transformer, it produces up to 1080p video with synchronized audio in a single pass. Atlas Cloud serves both HappyHorse 1.0 and 1.1 through one OpenAI-compatible key with pay-as-you-go pricing.
Both versions share the same architecture and the same text-to-video, image-to-video, and reference-to-video endpoints. HappyHorse 1.1 improves motion expressiveness, prompt adherence, and multi-reference fusion, so complex multi-shot narratives and brand-consistent clips come out more stable. HappyHorse 1.0 additionally offers a video-edit endpoint that transforms existing footage with text instructions.
Create an Atlas Cloud account, generate an API key from your dashboard, and send a request to any HappyHorse endpoint. Generation runs asynchronously: submit a prompt, receive a task ID, then poll the status endpoint until your video is ready. There is no subscription to set up since billing is pay-as-you-go per generation. Start building today.
Text-to-video, image-to-video, and reference-to-video are available on both HappyHorse 1.0 and 1.1, while 1.0 adds a video-edit mode for reworking existing footage. Every mode accepts prompts up to 2500 characters and returns video through the same task-based workflow, so switching between modes requires only changing the model path in your request.
Output renders at 720P or 1080P, with clip length adjustable from 3 to 15 seconds. Nine aspect ratios are supported, including 16:9, 9:16, 1:1, 4:3, 3:4, 4:5, 5:4, 21:9, and 9:21, covering everything from vertical social clips to ultrawide cinematic frames. A seed parameter also lets you reproduce a generation you like.
Need consistent characters or brand elements across a clip? Reference-to-video accepts one to nine reference images alongside a text prompt and keeps identity, product details, and style stable through the full video. Each image can be JPEG, JPG, PNG, or WEBP up to 20MB, with at least 400 pixels on the shorter side.
Yes. HappyHorse produces dialogue, ambient sound, and Foley effects together with the video in one forward pass instead of relying on a separate audio pipeline. Native lip sync covers seven languages: Mandarin, Cantonese, English, Japanese, Korean, German, and French, which makes it practical for multilingual campaigns without extra dubbing work.
Every HappyHorse API endpoint on Atlas Cloud is priced at $0.14 per generation, billed pay-as-you-go with no subscription or minimum commitment. Whether you call text-to-video on 1.0 or reference-to-video on 1.1, the per-call price stays the same, so costs remain predictable as your volume grows. Start today.
When HappyHorse 1.0 entered the Artificial Analysis Video Arena in April 2026, it immediately ranked #1 in both text-to-video and image-to-video based on blind human preference voting. In image-to-video without audio it reached an Elo of 1402, ahead of Seedance 2.0 at 1355 and Grok Imagine Video at 1331. Arena standings shift as new models launch, so check the live leaderboard for current rankings.
Join the Discord community for the latest model updates, prompts, and support.