Minimax H3 Workflow: Using Different Languages for Voice and Subtitles

aziib · reddit · 2026-08-07

A developer shared a practical workflow using the Minimax H3 model (ref2va workflow). By having Gemini read the model's documentation on HuggingFace and generate prompts, the workflow allows the character's voice to speak in Japanese while simultaneously displaying translated English subtitles. The author also included a link to a workflow optimized for low VRAM environments.

Original post →

More from Multimodal

Multimodal channel →