Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Introducing StoryMem — a memory-augmented framework for multi-shot long video storytelling. StoryMem carefully injects compact memory into the generation process with minimal overhead, enabling: • Cross-shot consistency • Smooth transitions • Narrative coherence across minutes-long videos Awesome work by Kaiwen Kaiwen Zhang (Kevin) Project: arXiv: Code:

Xingang Pan

3,409 subscribers

16,593 views • 9 months ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

2 Comments

James9 months ago

i hope they learn the ResNet lesson of residuals though.

Aaliya9 months ago

Impressive work! Maintaining consistency and narrative in long AI videos is a game-changer.

Related Videos

Meta presents OneStory Coherent Multi-Shot Video Generation with Adaptive Memory

Meta presents OneStory Coherent Multi-Shot Video Generation with Adaptive Memory

AK

11,532 views • 10 months ago

Following the release of Wan2.7-Image, Alibaba has launched Wan2.7-Video, its video generation model. By bringing together text, image, video, and audio inputs, addressing gaps in narrative coherence and multi-shot consistency. Watch the video and explore its capabilities!

Following the release of Wan2.7-Image, Alibaba has launched Wan2.7-Video, its video generation model. By bringing together text, image, video, and audio inputs, addressing gaps in narrative coherence and multi-shot consistency. Watch the video and explore its capabilities!

Alibaba Cloud

3,678,855 views • 5 months ago

Introducing Wan2.6 - A native multimodal model that turns your ideas into breathtaking videos and images! · Starring: Cast characters from reference videos into new scenes. Support human or human-like figures, enabling complex multi-person and human-object interactions with appearance and voice consistency. · Intelligent Multi-shot Narrative: Turn simple prompts into auto-storyboarded, multi-shot videos. Maintain visual consistency and upgrade storytelling from single shots to rich narratives. · Native A/V Sync: Generate multi-speaker dialogue with natural lip-sync and studio-quality audio. It doesn’t just look real - it sounds real. · Cinematic Quality: 15s 1080p HD generation with comprehensive upgrades to instruction adherence, motion physics, and aesthetic control. · Advanced Image Synthesis and Editing: Deliver cinematic photorealism with precise control over lens and lighting. Support multi-image referencing for commercial-grade consistency and faithful aesthetic transfer. · Storytelling with Structure: Generate interleaved texts and images powered by real-world knowledge and reasoning capabilities, enabling hierarchical and structured visual narratives.

Introducing Wan2.6 - A native multimodal model that turns your ideas into breathtaking videos and images! · Starring: Cast characters from reference videos into new scenes. Support human or human-like figures, enabling complex multi-person and human-object interactions with appearance and voice consistency. · Intelligent Multi-shot Narrative: Turn simple prompts into auto-storyboarded, multi-shot videos. Maintain visual consistency and upgrade storytelling from single shots to rich narratives. · Native A/V Sync: Generate multi-speaker dialogue with natural lip-sync and studio-quality audio. It doesn’t just look real - it sounds real. · Cinematic Quality: 15s 1080p HD generation with comprehensive upgrades to instruction adherence, motion physics, and aesthetic control. · Advanced Image Synthesis and Editing: Deliver cinematic photorealism with precise control over lens and lighting. Support multi-image referencing for commercial-grade consistency and faithful aesthetic transfer. · Storytelling with Structure: Generate interleaved texts and images powered by real-world knowledge and reasoning capabilities, enabling hierarchical and structured visual narratives.

Wan

3,848,218 views • 10 months ago

Wan2.6: Unlock multi‑shot storytelling with intelligent storyboarding, cross‑shot consistency, synchronized audio, and videos up to 15 seconds for richer expression. Build cinematic sequences from a single prompt. Follow us to stay updated on the latest Wan features!

Wan2.6: Unlock multi‑shot storytelling with intelligent storyboarding, cross‑shot consistency, synchronized audio, and videos up to 15 seconds for richer expression. Build cinematic sequences from a single prompt. Follow us to stay updated on the latest Wan features!

Alibaba Cloud

152,955 views • 10 months ago

HoloCine Holistic Generation of Cinematic Multi-Shot Long Video Narratives

HoloCine Holistic Generation of Cinematic Multi-Shot Long Video Narratives

AK

17,449 views • 1 year ago

Wan2.6：Multi-shot Storytelling & Cinematic Shot Control ✨ Intelligent Storyboarding: Generate coherent videos with multiple shots from a single prompt. 🎬 Cross-Shot Consistency: Maintains a consistent main subject and scene across different shots. 🔊 Synchronized Audio: Generates video and synchronized sound simultaneously for a complete experience. ⏱️ Up to 15s Duration: Create longer videos, allowing for more dynamic shots and richer expression. Try Wan2.6 now at

Wan2.6：Multi-shot Storytelling & Cinematic Shot Control ✨ Intelligent Storyboarding: Generate coherent videos with multiple shots from a single prompt. 🎬 Cross-Shot Consistency: Maintains a consistent main subject and scene across different shots. 🔊 Synchronized Audio: Generates video and synchronized sound simultaneously for a complete experience. ⏱️ Up to 15s Duration: Create longer videos, allowing for more dynamic shots and richer expression. Try Wan2.6 now at

Wan

18,066 views • 10 months ago

This is a step forward for long-context video tuning, which has already unlocked impressive multi-shot video generation pipelines from . Generating minutes- or even hour-long videos without drifting or forgetting is not "coming soon" -- it is already here. 3/4

This is a step forward for long-context video tuning, which has already unlocked impressive multi-shot video generation pipelines from . Generating minutes- or even hour-long videos without drifting or forgetting is not "coming soon" -- it is already here. 3/4

Gordon Wetzstein

27,473 views • 1 year ago

Introducing The Matrix --- a foundation world model for generating infinite-length, hyper-realistic videos with real-time, frame-level control: - Infinite-length video generation - 720p high-quality rendering - Real-time, frame-level control at 16 FPS - Generalization to real-world video control 🔗Blog: 📄Paper: 💻Code & Playable Demo: Coming soon! Key Innovation: A brand new technique called the shift-window denoise process model, enabling auto-regressive generation for diffusion and consistency models in real-time. Special thanks to project leader Ruili Feng and the entire Matrix team for their dedication and hard work over the year-long project.

Hongyang Zhang

178,482 views • 1 year ago

Introducing Amazon Nova Reel 1.1 🚀⚡🎬 Generate multi-shot videos up to 2-minutes in length & style consistency across shots. Available in #AmazonBedrock today. #AWS #generativeAI ▶️

Introducing Amazon Nova Reel 1.1 🚀⚡🎬 Generate multi-shot videos up to 2-minutes in length & style consistency across shots. Available in #AmazonBedrock today. #AWS #generativeAI ▶️

AWS AI

28,132 views • 1 year ago

🚀Seedance 2.0 is now open to EVERYONE, including US. Enjoy no queues & full power on Lovart. → Unmatched Length: 60-second video generation with Lovart Skills → Consistency Control: Auto three-view generation across long videos → Creative Prompts: Autonomous planning, no complex shot prompts needed → Smart Storyboarding: Seamless shot merging for cinematic transitions Plus, members enjoy 45% OFF + 150 FREE Seedance 2.0 Videos. Like + repost + follow – 10 lucky winners get a free month of Pro

🚀Seedance 2.0 is now open to EVERYONE, including US. Enjoy no queues & full power on Lovart. → Unmatched Length: 60-second video generation with Lovart Skills → Consistency Control: Auto three-view generation across long videos → Creative Prompts: Autonomous planning, no complex shot prompts needed → Smart Storyboarding: Seamless shot merging for cinematic transitions Plus, members enjoy 45% OFF + 150 FREE Seedance 2.0 Videos. Like + repost + follow – 10 lucky winners get a free month of Pro

Lovart AI

1,684,468 views • 6 months ago

PAI is officially rolling out. The long-form cinematic model from Utopai Studios, built for storytellers who think in scenes, not seconds. What PAI delivers: • Minutes-long continuous video generation • Character & scene consistency across every shot • Natural language editing — change the story, not just the frame Watch the first story created entirely with PAI.

PAI is officially rolling out. The long-form cinematic model from Utopai Studios, built for storytellers who think in scenes, not seconds. What PAI delivers: • Minutes-long continuous video generation • Character & scene consistency across every shot • Natural language editing — change the story, not just the frame Watch the first story created entirely with PAI.

Utopai Studios

4,121,791 views • 7 months ago

We propose Long Context Tuning (LCT) for scene-level video generation to bridge the gap between current single-shot generation and real-world narrative video productions. Homepage: Report:

We propose Long Context Tuning (LCT) for scene-level video generation to bridge the gap between current single-shot generation and real-world narrative video productions. Homepage: Report:

Ceyuan Yang

46,908 views • 1 year ago

New research with Tsinghua University: Spatial-TTT. A framework for streaming visual-based spatial intelligence with test-time training (TTT). Spatial-TTT adapts fast weights to capture and organize spatial evidence from long video streams, enabling models to build structured 3D spatial memory over time. Highlights: 🔹Efficient streaming memory. Fast weights act as compact spatial memory with sublinear memory growth over 7000+ frames and more than 40% lower compute. 🔹Spatial-predictive mechanism. TTT layers with 3D spatiotemporal convolution capture geometric correspondence and temporal continuity. 🔹SOTA results on long-horizon video spatial understanding (VSI-Bench). The paper ranked #1 on Hugging Face Daily Papers on March 13. Project page: GitHub: Paper: Model & Data:

New research with Tsinghua University: Spatial-TTT. A framework for streaming visual-based spatial intelligence with test-time training (TTT). Spatial-TTT adapts fast weights to capture and organize spatial evidence from long video streams, enabling models to build structured 3D spatial memory over time. Highlights: 🔹Efficient streaming memory. Fast weights act as compact spatial memory with sublinear memory growth over 7000+ frames and more than 40% lower compute. 🔹Spatial-predictive mechanism. TTT layers with 3D spatiotemporal convolution capture geometric correspondence and temporal continuity. 🔹SOTA results on long-horizon video spatial understanding (VSI-Bench). The paper ranked #1 on Hugging Face Daily Papers on March 13. Project page: GitHub: Paper: Model & Data:

Tencent Hy

20,792 views • 7 months ago

［100625］ Kaiwen performing as a backup dancer for Lay's concert last night! And it was a big opportunity to Kaiwen perform in a big stage. We are so proud of you Kaiwen🥹 Anyways Happy birthday to the King of China Lay Zhang 💕 #CHENKAIWEN #천카이원 #陈凯文 #チェンカイウェン

［100625］ Kaiwen performing as a backup dancer for Lay's concert last night! And it was a big opportunity to Kaiwen perform in a big stage. We are so proud of you Kaiwen🥹 Anyways Happy birthday to the King of China Lay Zhang 💕 #CHENKAIWEN #천카이원 #陈凯文 #チェンカイウェン

CHEN KAIWEN GLOBAL

64,307 views • 1 year ago

MuseSteamer just leveled up! Our video generation model now supports real-time interactive long-form video generation. It breaks the traditional 10-second limit, creating videos of any length with greater speed and control—enabling users to pause, rewrite storylines, or extend transitions any time during the process.

MuseSteamer just leveled up! Our video generation model now supports real-time interactive long-form video generation. It breaks the traditional 10-second limit, creating videos of any length with greater speed and control—enabling users to pause, rewrite storylines, or extend transitions any time during the process.

Baidu Inc.

142,732 views • 1 year ago

Now autoregressive video models can remember. 🧠 Introducing MosaiChunk: persistent visual memory for long-horizon autoregressive video generation🧩. Instead of remembering everything, it remembers the right things. ✍️ 1/6

Now autoregressive video models can remember. 🧠 Introducing MosaiChunk: persistent visual memory for long-horizon autoregressive video generation🧩. Instead of remembering everything, it remembers the right things. ✍️ 1/6

Haven Feng

31,008 views • 12 days ago

Introducing LongLive 2.0 by @NVIDIA. Real-time text-to-video with full multi-shot control. Build your story shot by shot, scene by scene. Available now on Reactor.

Introducing LongLive 2.0 by @NVIDIA. Real-time text-to-video with full multi-shot control. Build your story shot by shot, scene by scene. Available now on Reactor.

reactor

187,752 views • 4 months ago

Introducing Long-term, multiplayer memory for your agents. #1 on BEAM, the largest public memory benchmark. Install it with a single prompt. Keep your stack. Add a past. Try it today →

Introducing Long-term, multiplayer memory for your agents. #1 on BEAM, the largest public memory benchmark. Install it with a single prompt. Keep your stack. Add a past. Try it today →

Mehdi Djabri

1,090,794 views • 10 days ago

Introducing DuoAttention: Our new framework slashes both memory and latency for long-context LLMs without sacrificing performance! By applying full KV cache only to critical heads, we achieve: ⚡ 2.55x memory reduction ⚡ 2.18x decoding speedup ⚡ 3.3M tokens on a single A100 GPU

Introducing DuoAttention: Our new framework slashes both memory and latency for long-context LLMs without sacrificing performance! By applying full KV cache only to critical heads, we achieve: ⚡ 2.55x memory reduction ⚡ 2.18x decoding speedup ⚡ 3.3M tokens on a single A100 GPU

Guangxuan Xiao

31,108 views • 2 years ago

Memory-V2V by Adobe: diffusion model with memory for consistent iterative editing; - ReCamMaster+LucyEdit; - reduces FLOPs by 30-90%; outperforming FIFO-Diffusion. high fidelity across long durations like consistent style transfer

Memory-V2V by Adobe: diffusion model with memory for consistent iterative editing; - ReCamMaster+LucyEdit; - reduces FLOPs by 30-90%; outperforming FIFO-Diffusion. high fidelity across long durations like consistent style transfer

Wildminder

10,248 views • 9 months ago