MiniMax H3 keeps adding unwanted voices and music despite prompts saying no

bickid · reddit · 2026-11-17

Reddit user bickid reports that videos generated with MiniMax H3 look great, but the audio track keeps going rogue: the model inserts voices or background music never requested — and sometimes directly contradicts the prompt, ignoring explicit instructions like "no background music, nobody talks."

Someone previously linked a website with a workaround, but he couldn't make sense of it and is asking MiniMax H3 users for a simple, plain-language explanation (with examples) of how to suppress unwanted music and voices. The thread highlights a clear audio instruction-following weakness in the video model.

Original post →

More from Multimodal

Multimodal channel →