Not the VFX team. Sound.
This week, a new AI model called MiniMax H3 launched out of China. It doesn't just generate video, it generates the voice, sound effects, and music in the same pass, as real stereo audio. Here's what that actually means,
One Render, Not Three Departments
Up to 15 seconds, 2K resolution, and MiniMax's own pricing claim: less than a third the cost of "mainstream" AI video models. That's not a video tool with an audio bonus. That's a foley artist, a composer, and a mixer's job folded into one render call.
Sound Was Supposed to Be Safe
We've spent months talking about AI coming for VFX and editing. Sound's mostly been treated as safe, a craft with real ears behind it a model can't fake. This is the first release built specifically to close that gap, not just improve the picture.
An Honest "Not Yet"
Is it as good as a real mix from someone who's spent years on a dub stage? I haven't heard it against a real project yet, and neither has anyone reviewing it this week. That's a fair, honest "not yet", not a "never."
The Uncomfortable Pattern
"Not as good yet" has been true of every AI tool that ended up mattering, right up until it wasn't. That's not fear-mongering, it's just the actual track record so far.
My Honest Take
If you work in sound (foley, ADR, mixing, scoring) this is worth ten minutes of your attention this week. Not because it's ready to replace you. Because it's the first one actually built to try.
Has This Come Up in Your Sound Department Yet?
A new AI model launched this week generating video plus its own voice, sound effects, and music, in one render, as real stereo audio.