MiniMax-H3: Omni-Modal Model Generates Video with Audio on Apple Silicon

MiniMax-H3 is an omni-modal generative system from MiniMax that accepts text, images, audio, and video, and generates video clips up to 15 seconds with embedded audio.

Thanks to the PipeNetwork/minimax-h3-mlx package based on the MLX framework, the model runs on Apple Silicon, but requires significant memory and long generation times.

Yes, but the high computational cost makes it more suitable for experiments and research than large-scale production, especially in environments without strong infrastructure.