Alibaba HappyHorse 1.0 R2V

Reference-to-video generation with identity-consistent subjects and synchronized native audio

About model

HappyHorse 1.0 is the video generation model from Alibaba's ATH AI Innovation Unit, built to generate video and synchronized audio together in a single pass. This reference-to-video endpoint generates from reference inputs, keeping characters and products consistent across the clip while adding motion, dialogue, ambient sound, and effects aligned to the on-screen action. At launch the model topped both the text-to-video and image-to-video Elo rankings on the Artificial Analysis Video Arena, where results come from blind human preference votes. Prompts are processed natively in multiple languages, including English, Chinese, and Japanese. Available on Together AI.

‍

Video Arena at Launch (Artificial Analysis)

Topped the text-to-video and image-to-video Elo rankings

Audio Generation

Native

Dialogue, ambient sound, and effects in the same pass as the video

Output Resolution

Up to 1080p

Cinematic output with physics-consistent motion

Model key capabilities

Identity Consistency: Generates from reference inputs, keeping characters and products stable across the clip
Native Audio Generation: Dialogue, ambient sound, and effects generated in the same pass and aligned to on-screen action
Leaderboard-Ranked Quality: Topped the Artificial Analysis Video Arena text-to-video and image-to-video Elo rankings at launch
Production-Ready Infrastructure: 99.9% SLA, available on Together AI serverless infrastructure

API usage

cURL
Python
Typescript

Endpoint:

alibaba/happyhorse-1.0-r2v

curl --request POST \
  --url https://api.together.xyz/v2/videos \
  --header "Authorization: Bearer $TOGETHER_API_KEY" \
  --header "Content-Type: application/json" \
  --data '{
    "model": "alibaba/happyhorse-1.0-r2v",
    "prompt": "some penguins building a snowman"
  }'

from together import Together

client = Together()

# Create a video generation job
job = client.videos.create(
    prompt="A serene sunset over the ocean with gentle waves",
    model="alibaba/happyhorse-1.0-r2v"
)

import Together from "together-ai";

const together = new Together();

async function main() {
  // Create a video generation job
  const job = await together.videos.create({
    prompt: "A serene sunset over the ocean with gentle waves",
    model: "alibaba/happyhorse-1.0-r2v"
  });

Model card
Architecture Overview:
• Unified generation of video and synchronized audio in a single pass, with no separate audio pipeline
• Motion engine designed around real-world physical constraints to reduce warping artifacts
• Native multilingual prompt processing, including English, Chinese, and Japanese

Training Methodology:
• Developed by Alibaba's ATH AI Innovation Unit (formerly the Taotian Future Life Lab)
• Debuted anonymously on the Artificial Analysis Video Arena before Alibaba confirmed it as the developer

Performance Characteristics:
• Topped the Artificial Analysis Video Arena text-to-video and image-to-video Elo rankings at launch (April 2026), based on blind human preference votes
• Elo scores in the with-audio tracks placed among the leading models on the leaderboard
‍
Prompting
Together AI API Access:
• Access HappyHorse 1.0 R2V via Together AI APIs using the endpoint alibaba/happyhorse-1.0-r2v
• Authenticate using your Together AI API key in request headers
• Provide reference inputs with a text prompt to keep subjects consistent, and poll the video response until generation completes
• Prompts can be written natively in multiple languages, including English, Chinese, and Japanese
• Available on Together AI serverless infrastructure
‍
Applications & use cases
E-Commerce & Product Video:
• Turn product imagery into showcase clips with matched sound effects by calling the Together endpoint
• Generate consistent brand visuals across multiple video variations for storefronts and listings
• Localize product videos with prompts written natively in English, Chinese, or Japanese

Marketing & Social Content:
• Produce short-form ads and social clips with dialogue and ambience generated in one pass
• Iterate on campaign concepts quickly without a separate audio or dubbing step
• Batch-generate platform-ready variations through the Together API

Narrative & Creative Production:
• Draft storyboards and previsualization clips with physics-consistent motion
• Build character-driven scenes with synchronized dialogue and sound
• Prototype creative directions before committing to full production
‍

Related models

Model specifications

Model data

Model provider
Alibaba
Type
Video
Resolution/Duration
Up to 1080p
Deployment
Serverless
Endpoint
alibaba/happyhorse-1.0-r2v
Input modalities
Text
Image
Output modalities
Video

Category
Video

Run in Playground

Quickstart docs

Deploy model

Alibaba HappyHorse 1.0 R2V

About model

API usage

Model card

Prompting

Applications & use cases