WAN3.0 Omni-Reference generates videos directly from documents without prompts
Aiden_Tech_Ai · x · 2026-08-20
Video model WAN3.0 demonstrated a feature called Omni-Reference. It can ingest a slide deck, document, sheet, or webpage and generate a video based on it, requiring no additional prompt or storyboard. The demo showed a simulated advertisement video created directly from a presentation deck.
More from Multimodal
- Largest Open Human Image Preference Dataset Released: 2M+ Annotations, 30 Models Ranked, $1M Grant — _akhaliq · 2026-08-20
- Fidget: Local AI music generation tool for low-end GPUs — D3VAUX · 2026-08-20
- Runway gen-2 video: Ghosts dreaming of lives unlived — c_valenzuelab · 2026-08-20
- Qwen response times 4x faster with prompt tweaks only — iamrobotbear · 2026-08-20
- Open-Source AI Roundup: Qwen Video Edit, LTX 2.5, and Practical Benchmarks — fromourback · 2026-08-20
- Creating J-Pop/K-Pop MVs with H3, SUNO, and AudioLock — xyzdist · 2026-08-20