Open-source VDN speeds up MiniMax-H3 video 75-90x, generating 14s of 768p in 11s on 8xB200
AiCreatorCamp · reddit · 2026-09-03
Developer haochengxiucb released Video Delta Net (VDN), an open-source hybrid-attention method for near-lossless live text-to-video.
- Accelerates MiniMax-H3 video generation by 75-90x
- Generates 14 seconds of 768p video in 11 seconds on 8x NVIDIA B200 GPUs — faster than playback with minimal quality loss
- Weights are on Hugging Face (OpenVDN/vdn-minimax-h3); the author is waiting for the ComfyUI team to integrate optimizations
Related event: Open-Source Video Delta Net Pushes Video Generation Faster Than Playback(8 posts)→
More from Multimodal
- Ex-Brain Games host made a 25-min AI film in 5 days using 'vibe directing' in iMessage — lmoroney · 2026-10-21
- Motion Launches a General Video Agent That Edits Footage and Mimics Any Reference Style — mhdfaran · 2026-10-21
- Ant open-sources 5 Ming-3.0 speech models spanning TTS, audio understanding and ASR — Med1_Ai · 2026-10-21
- What keeps AI videos watchable: hooks, consistency and a payoff — aftahi_ai · 2026-10-21
- MORPHJUTSU skill in Flova transfers any video's motion to your character — aftahi_ai · 2026-10-21
- NVIDIA open-sources MotionBricks: 350K motion skills at 15,000 FPS, 2ms latency — aftahi_ai · 2026-10-21