Modèles temps réel
MiniMax H3
Modèle vidéo 33B open-weight (Hailuo 3.0) — clips de 15 s avec audio stéréo natif ; la base de cette vague.
huggingface.co · répertorié ≠ recommandé · revérifié le 2026-09-03
- Paramètres
- 33B dense · open base weights (H3-Base)
- Sortie
- 4–15s clips; 2K via API only (local ≈768p)
- Audio
- Native stereo — a property of the model
- Poids ouverts
- Aug 3, 2026
- Note de licence
- Regional exclusions incl. US / EU / UK / KR
Qu’est-ce que MiniMax H3 ?
Le modèle vidéo open-weight de 33B (Hailuo 3.0) sur lequel repose cette vague. Seul le base est ouvert (≈768p en local) ; le pass 2K et les modules de référence sont via API, et la licence exclut plusieurs régions. L’API officielle de MiniMax n’est pas temps réel — c’est la version realtime de fal qui en fait un support de streaming.
Ce que fait MiniMax H3
- 33B dense open-weight model (H3-Base) released Aug 3, 2026
- Generates 4–15 second clips with native 32kHz stereo audio — audio is in the model, not post-processed
- Omni-modal: text, image and video in; video and audio out
- 2K upscale and context-reference modules are API-only — local open-weight generation is around 768p
- Day-zero ecosystem support from Hugging Face, ComfyUI, ModelScope and many chip vendors
À qui s’adresse MiniMax H3 ?
- Self-hosting open-weight video generation on a local GPU
- Building video pipelines that need native, controllable audio
- Running realtime or near-realtime generation via fal’s H3 Max build
Points forts et limites de MiniMax H3
Points forts
- One of the first flagship open-weight video models with native audio
- Runs locally — ComfyUI shipped day-zero support and cites RTX-3060-class GPUs
- Day-one adaptation across major open-source tooling and silicon vendors
Limites
- Only the base is open; the 2K pass and reference modules are API-only
- Licence carries regional exclusions (incl. US, EU, UK, South Korea)
- MiniMax’s own API is not realtime — the streaming use case runs through fal’s tuned build
Ce qu’on en dit
Citations issues de couverture publique — collectées le 2026-09-03.
“MiniMax H3 dropped today with open weights, and it’s natively supported in ComfyUI as of this morning. Day zero.”
“Audio is a property of the model, not a post-process. Every audio output is native stereo.”
Questions fréquentes
Is MiniMax H3 open source?
The H3 base weights are open (Aug 3, 2026); the 2K regeneration and context-reference modules are API-only.
Can I run MiniMax H3 locally?
Yes for the open base — around 768p resolution; 2K output needs the API modules.
What does native audio mean for H3?
Audio is generated as a property of the model — native stereo — rather than added as a post-processing step.
Is MiniMax H3 realtime?
Not on MiniMax’s official API. fal’s H3 Max build post-trains it to generate faster than playback for live use.