MiniMax H3: video that arrives with its own soundtrack, self-hosted
中文摘要
MiniMax H3 是一个 330 亿参数的全能 Transformer 模型，可在单一潜空间流中同时生成视频与立体声。
English Summary
MiniMax H3 is a 33B parameter omni-transformer that simultaneously predicts video and stereo sound within a single latent stream.
Original Excerpt
A 33-billion-parameter omni-transformer that predicts picture and stereo sound in the same latent stream — running on a DGX Spark in our… Continue reading on Medium »