Real-time full-duplex interactive role model
Vivix is a platform dedicated to developing real-time interactive multimodal models, featuring two primary models. Vivix-A1 is touted as the first real-time full-duplex model capable of creating interactive AI characters that can move their entire body, listen, speak, and respond in real time. These characters can be interrupted at any moment during conversation, with audio and visual outputs remaining synchronized – ensuring that there is no delay of more than 0.6 seconds between the initial input and the corresponding response.
Vivix-W1 is a streaming, native multimodal model designed to generate open-world environments in real time, allowing for free exploration by users.
Its application scenarios include live-streaming e-commerce, virtual companionship, interactive entertainment, gaming, and customer service guidance. The API has been made available through platform.vivix.ai on a pay-per-use basis. According to the official statements, the inference costs are less than 1 dollar per hour; however, the specific price per use is not disclosed on the official website. This solution is more suitable for developers and enterprises to integrate into their systems, while individual users have limited opportunities to directly utilize its functions.
Before using it to create vertical-screen short dramas, you can check in TVLab how popular shows on the same topic begin and what twists occur in which episodes. Then, select an inspiration card that already contains a detailed storyboard and use it as a guide to generate each scene one by one.