Step Stars Collaborates with Geely to Open Source the Step Series Multimodal Large Models for Video and Audio Fields

Stairway Stars and Geely Automotive Group have jointly announced that the Step series of multimodal large models, co-developed by both parties, will be open-sourced globally for developers. The open-source models include two versions: the video generation model Step-Video-T2V and the voice interaction model Step-Audio.

Step-Video-T2V is a video generation model with 30 billion parameters, capable of generating high-quality videos at 204 frames and 540P resolution. According to Stairway Stars, this model leads in parameter count and performance among currently available open-source video generation models.

The other open-source model, Step-Audio, is the first product-grade open-source voice interaction model in the industry. It can generate voice expressions with emotion, dialects, languages, singing, and personalized styles based on different scenario requirements, enabling high-quality conversations with users. Moreover, the voices generated by Step-Audio are natural and emotionally intelligent, and it supports high-quality voice replication functionality.

Users can now experience the new features of Step-Audio through the Yuewen App provided by Stairway Stars. This open-source initiative marks a further deepening of cooperation between Stairway Stars and Geely Automotive Group in the field of artificial intelligence, and provides global developers with more possibilities for innovation.