Omni 1.1 adds video references: 3-second clips ensure character consistency in AI video
fofrAI · x · 2026-09-04
fofrAI demonstrates Omni 1.1's new video reference feature: short clips can be used as references to insert consistent, accurate characters and objects into generated videos. In the demo, two 3-second reference clips of a handstand were combined with reference syntax and a prompt to generate "she is doing a handstand on the car while it is being driven." Video extension can also be combined with video references for tighter control.
Related event: Omni 1.1 Adds Video Reference for Consistent Characters and Mechanisms(2 posts)→
More from Multimodal
- Astra one-shots a historically accurate 3D reconstruction of Waterloo — danshipper · 2026-09-04
- Tekken 3's Nina vs Leo remade with AI in cinematic realism — SimplyAnnisa · 2026-09-04
- Digital abstract expressionism made with AI — floguo · 2026-09-04
- GPT-6 Astra Demolishes Fable 5.1 in Scene Generation Demo, Claims 'Astra Is AGI' — VraserX · 2026-09-04
- MiniMax character LoRA on Runpod: 150 images, ~2300 steps to nail likeness, and 5090 beats H100 — Draufgaenger · 2026-09-04
- AI-generated video of a celestial entity crushing an army goes viral — eyishazyer · 2026-09-04