Structure raw media into preliminary highlights before the manual edit starts.
Up to 1,500 tokens/sec
Private local AI for long-context reasoning and fast inference.
