MiniMax H3 voice reference reportedly broken: ref2va only works as direct pass-through
Beneficial_Fish_7509 · reddit · 2026-10-25
A Reddit user reports that MiniMax H3's ref2va workflow ignores reference audio: instead of cloning Voice A onto a new sentence, it generates a random unrelated voice. After a full day of single-variable troubleshooting (prompts, formats, turbo loras, timing), the only 'working' case is when the reference audio says the exact same sentence — i.e., pass-through. Reference images work fine, and other users have reproduced the issue in an open Hugging Face discussion.
More from Multimodal
- Higgsfield lands NVIDIA and TCV in Series B extension, crosses $1B annualized revenue — aakashgupta · 2026-10-25
- enGEN demo shows AI video with oil-painting cinematic look — jordanbayne · 2026-10-25
- Indie dev releases Visage-Image face T2I model; custom narrow-domain VAE beats FLUX.2 on iFID — Sherstnyov · 2026-10-25
- Stunning AI-generated street drive video looks near-photoreal end to end — Extraaltodeus · 2026-10-25
- AI animation flood incoming: Three.js slop now, Claude's telltale editing sounds — ChrisGPT · 2026-10-25
- Ecco the Dolphin Reimagined with AI Video — demondisc · 2026-10-25