Use case: a music video made fully locally with MiniMax H3, YuE2 and HiDream O1
A creator made a three-minute music video using only open-weight models on their own machine: YuE2 for the song, HiDream O1 for character and world references, Qwen 3.8 27B to format prompts and MiniMax H3 for the video. The edit was done in DaVinci Resolve with only basic tools and no cloud AI.
What it does
Each stage runs locally: YuE2 makes the music, HiDream O1 makes the character and world reference images, Qwen 3.8 27B turns ideas into formatted prompts, and MiniMax H3 animates the shots. The cut is put together in DaVinci Resolve using only basic, non-cloud editing tools.
Results and limits
The author calls the all-local route a pain: when something breaks, only the open-source community can help, and errors are often so niche nobody else has hit them. In return they control the exact parameters, seed, LoRA stack, fine-tune, custom nodes, workflow and hardware, so they can recreate any shot and aren't exposed to hosted models changing, censorship or account bans.
Worth knowing
Check the licenses before commercial work: YuE2-3B's weights are CC BY-NC 4.0 and HiDream O1 Image is MIT. For MiniMax H3: Users in the EU, UK, US and South Korea need to request a license from MiniMax via the H3 License Request Form (platform.minimax.io/h3-license) before use.