MLX-Serve Adds Video Storyboards, Music Tools and Linux NVIDIA Builds
Version 26.10.2 expands the local inference server into a broader media studio and ships Linux packages for NVIDIA machines.
A wider local AI studio
The mlx-serve team released version 26.10.2 on October 12, adding video storyboards, music and sound-effect tools to its local inference server. The release notes say users can turn a story into a sequence of shots joined as a clip, generate songs with YuE2, and make sound effects with Stable Audio 3. A built-in web console now groups chat, image, audio and video workflows with a library and server monitoring. These are the developer’s descriptions of shipped features, not independent tests of output quality or speed.
Beyond the Mac
The release also offers Linux downloads for x64 and arm64 NVIDIA systems. Its English README specifies an NVIDIA driver version of at least 580, CUDA 13 and other libraries, so the Linux package is not a general CPU-only install. The team says the server supports recent open models and GGUF files while allowing agent chats to continue during media generation. It reports faster inference and heavy-load stability from its own test setups. Those performance and reliability figures have not been independently verified here, and usable models will still depend on a machine’s memory and GPU.