MiniMax H3 Max Reference-to-Video Goes GA on fal, RTF Drops to 0.876
isidentical · x · 2026-09-06
fal announces that Reference-to-video on MiniMax H3 Max is now generally available, up to 2X faster with significantly higher quality:
- RTF down from 1.5 to 0.876, enabling real-time generation with up to 4 references
- Better context understanding: references and prompt are automatically aligned in the same semantic domain
- Much stronger reference preservation than both the preview and the original H3, thanks to continued RL on character consistency and other tasks
The feature previously shipped as an early preview with RTF around 1 at 768p (1.5 with references).
More from Multimodal
- GPT-6 Astra Wows Users With 3D Generation, Front-End Coding Still Lagging — CtrlAltDwayne · 2026-09-06
- Rebuilding Konoha from Naruto in 3D with Higgsfield + GPT-6 Astra in 10-15 minutes — SimplyAnnisa · 2026-09-06
- Lyria 3.5 doubles as a TTS model: one prompt yields eerily natural 60-second speech — fofrAI · 2026-09-06
- A Mechanical Dream Engine Built with GPT-6 Astra Spanning Four Worlds — techartist_ · 2026-09-06
- Image + video + agents combined: new step in agentic video creation opens for feedback — Lianhuiq · 2026-09-06
- Overworld AI jokes its model shows an 'emergent lack of physical understanding' — lcastricato · 2026-09-06