MiniMax launched Hub, a multimodal AI video generator that merges image creation, video, voiceover, music, and editing into a single platform, at the Shanghai International Film Festival. The tool shifts the creative workflow from juggling five separate apps to working with one AI agent-while keeping a human firmly in control of every key decision.
A single agent, not a toolbox
"We used to open one AI tool for images, another for video, then others for voiceover, music, and finally a video editor to piece it all together," said Xu Lüyang from MiniMax's product operations department. "Now, we simply tell Hub our goal in natural language, or drop in a PDF proposal, a reference video, or an asset pack."
The agent inside Hub parses that input, breaks the project into tasks, picks the right models, runs them, and checks quality. Creators only need to step in to review and refine at critical moments. The result is a tool that operates less like a smarter generation button and more like a project runner that understands asset packs and proposal documents.
Among Generative Video platforms, Hub's built-in multimodal pipeline removes the need to export files between separate image, voice, and editing tools. MiniMax's own video generator, Hailuo AI, already handled video generation; Hub now layers orchestration on top.
Decision pauses, not a one-click black box
The system deliberately pauses at every major decision point rather than auto-generating a final product. "The AI agent shouldn't be a black box, one-click generator," Xu said. "It would pause at every critical decision point for your confirmation. We believe AI should lift the execution burden off human shoulders, but the creative direction and aesthetic judgment must ultimately be left to humans."
That human-in-the-loop design extends to how the tool learns from its user. Hub includes Skill and Memory functions that let creators hand over their workflows, aesthetic standards, and prompt engineering patterns. "We can hand over our workflows, aesthetic standards, and prompt engineering expertise to the agent, letting it memorize them so it can execute the tasks on its own next time," Xu said. The memory carries forward across projects, so the agent adapts to a specific creative voice rather than defaulting to generic outputs.
Real-world testing with short-film creators
MiniMax is collaborating with AI Backlot, Shanghai's AI work lab, by giving four creator pairs from the current cohort access to Hub for producing their short films. This places the tool directly into production environments where creative teams work under real deadlines. Pan Yuying, MiniMax Multimodal's head of video content, also demonstrated document generation across multiple file formats during the event, signaling that Hub handles more than just video files.
Why this matters for creatives
Hub targets the practical friction that slows down video production: tool-switching, file juggling, and re-explaining creative intent to each separate app. For creatives, the difference between a good tool and a great one often comes down to how little it interrupts the act of making decisions. By consolidating execution layers behind a human-in-the-loop checkpoint system, MiniMax is betting that the fastest AI workflow isn't the one that generates everything automatically-it's the one that learns when to step aside and let a creative call the shot. Resources that track AI for Creatives show a rising demand for tools that respect that boundary. The real test will be whether Hub's agent can handle the messy, non-linear way creative projects actually unfold.
Your membership also unlocks:
