happyhorse-1.0-t2v

HappyHorse-1.0-T2V is a Vincent video variant that takes only text as input and directly generates video content with synchronized audio. The model has "highly restored dynamic picture generation capabilities", can accurately understand text semantics, and output smooth, natural, high-quality videos with rich details. Like I2V, T2V’s inference service is provided by Alibaba Cloud Bailian and supports global multi-region deployment, with the same RPM of 300. In Artificial Analysis’ Vincent Video Arena, the T2V track topped the list with an Elo score of 1333, surpassing popular models such as ByteDance Seedance 2.0. The model natively supports lip synchronization in seven languages: English, Mandarin, Cantonese, Japanese, Korean, German, and French. It takes about 38 seconds to generate a 5-second 1080p video on a single H100.

Context

0 tokens

Max output

0 tokens

Input

llm image video

Output

video