Minimax H3 Ref2VA Test: Multi-Reference Audio Generation, Background Sound Hard to Control
aziib · reddit · 2026-08-06
A Reddit user shares experience with Minimax H3 Ref2VA for multi-reference audio generation. Using 2 reference images and 2 reference audio clips, generation takes about 18 minutes. It can generate sound effects, but background ambient sounds (like city hum, crowd, rain) are not output, and the voice is too clean. The user asks how to make background sounds louder. Hardware: RTX 4060 Ti 16GB + 64GB RAM. Workflow available on Civitai.
More from Multimodal
- Why Your AI Video Broke: 6 Common Failure Patterns and Fixes — Choice_Volume4090 · 2026-08-06
- Testing Hailuo: Generating Emotional Video with Regional Accents — Disastrous_Coast7870 · 2026-08-06
- Seeking Midjourney Alternatives That Prioritize Creativity Over Photorealism — Vozka · 2026-08-06
- Midjourney Test: Using SREF Code to Create Tense, Dark Supernatural Art — tisch_eins · 2026-08-06
- ComfyUI Video Generation Hurdle: Forced Speech Censorship Ruins Output — YouNoTypey · 2026-08-06
- Suno Unveils Responsible AI Music Principles and Transparency Tools — suno · 2026-08-06