MiniMax H3 Ignores Voice Volume Prompts; audio_reuse Offers a Partial Workaround
Dogluvr2905 · reddit · 2026-08-21
A user testing MiniMax H3 found that generated voices stay loud regardless of prompts like "quiet" or "distant mic," or even drastically reduced gain on the reference wav—H3 apparently reads waveform patterns, not volume. A workaround is audioreuse with a pre-lowered wav, but it replaces ALL audio in the clip, killing environmental sounds.
More from Multimodal
- fal launches Worlds for real-time creation and interaction with AI-generated worlds — Scobleizer · 2026-10-16
- Gummy worms as a physics test: Gemini 4 Argon vs. Claude Opus 5.5 elasticity sims — DynamicWebPaige · 2026-10-16
- 260M-parameter looped diffusion transformer beats 6x larger baselines with 5x less compute — burkov · 2026-10-16
- Seedance 2.5 demo: stretching one second of an explosion into a 30-second shot — LudovicCreator · 2026-10-16
- Octopus Short Film Made with Grok Imagine, Upscaled via Topaz — creatoroff · 2026-10-16
- Open-sourced 61-second computing history film generated entirely by Muse Spark — sophiamyang · 2026-10-16