← sources

Regarding Minimax Ref2Vid, how have you handled the characters' voices?

source: reddit (reddit:StableDiffusion)published 2026-11-06 18:52ingested 2026-11-07 02:39

retained text





I’ve been trying to create scenes featuring a couple of characters for a while now, but I haven’t found a solution yet—so I’m turning to those with more experience for some advice.



If there’s only one character in a scene, there’s no issue: if I upload an image and a reference audio clip, Minimax perfectly generates a video featuring that character with the corresponding voice.



However, if two characters appear in the video, Minimax struggles to distinguish which voice belongs to which character. For instance, if Character A speaks first, followed by Character B, the output isn't correct—even when I specify that Character A should have "Voice A" and Character B should have "Voice B."



I’ve tried using AI models like Gemma—uploading the official Minimax Ref2Vid guide as a PDF—but I still haven’t had any success.



Does anyone know how to resolve this? It’s a shame to have such great-looking videos but be unable to correctly assign the different audio tracks.



Even just an example prompt would be a huge help, so I could study it and figure things out.



Thanks to anyone who can help!

document fetched 2026-11-07 00:29 · canonical https://reddit.com/r/StableDiffusion/comments/1xpkvth/regarding_minimax_ref2vid_how_have_you_handled