MiniMax opens H3 on Hugging Face as a 33B multimodal video model with stereo audio
mark_k · x · 2026-08-04
MiniMax releases H3, an open-weights multimodal video model with stereo audio
MiniMax says H3 is now open on Hugging Face. It is a general-purpose omni-modal generative model that unifies text, images, video, and audio in one context.
- Generates video with native stereo audio and supports outputs up to 15 seconds at 2K.
- The company highlights strong instruction following, accurate text and brand rendering, and video-to-video motion transfer.
- It is aimed at advertising, branding, e-commerce, product design, and gaming.
- The post says H3 is a 33B-parameter Omni-Transformer conditioned on Qwen3-VL-32B.
- Checkpoints listed: FL2VA and Ref2VA; local inference works with Diffusers, SGLang, vLLM, and native ComfyUI support.
Related event: MiniMax H3 Local Tests: Benchmark Results and Comparisons Across GPUs(88 posts)→
More from Multimodal
- Fan animation made in Blender with Claude Code (Opus 5.5) — Icy-Appointment3583 · 2026-10-16
- Making Blender animations with Opus, Astra and Argon: how you prompt still matters — chongdashu · 2026-10-16
- Tessellation-free geometric queries on displaced surfaces cut caustics variance 10x — ssh4net · 2026-10-16
- Complex multi-link CAD mechanism finally animates, dev celebrates the breakthrough — zetalyrae · 2026-10-16
- OpenMOSS releases MOSS-Image-SFT: Apache 2.0 text-to-image model on Hugging Face — bdsqlsz · 2026-10-16
- Hyperflow open-sourced: 49 sampling steps down to 8, ~3x faster video generation — mhdfaran · 2026-10-16