Xiaodong Cun

Assistant Professor @ Great Bay University ||
Here's to the Crazy Ones. || Animating pixels. || Computer Vision/Graphics Researcher

[一然 王 @wang_yi75406]

🎮 We introduce GameHorizon Suite, a large-scale data and evaluation suite spanning multiple horizons and AAA games. It serves as a standardized yardstick for VLMs, UMMs, GUI, coding, and game agents.

Github: https://t.co/Q3MS8A53eT
HF Daily: https://t.co/GMw7gJ7wl0
Project Page: https://t.co/bN5JpvDWsA

[Ying Shan @yshan2u]

Introducing GameHorizon: a unified benchmark for AI gameplay across games, model families and task scales.

• Challenge: Existing benchmarks often cover few games, lack language instructions or rely on high-variance online rollouts, making consistent model comparison difficult.

• What: A unified data and evaluation suite spanning multiple temporal horizons, diverse AAA games and a broad range of model families.

• How: Aligning videos and actions with automatically annotated operations, goals and strategies—combined with reproducible offline evaluation and stepwise online testing.

• Result: Evaluation of 47 models reveals that future-action planning and goal decomposition remain key bottlenecks. Among the 12 models also tested online, stronger offline performance generally corresponds to better gameplay results.

[DailyPapers @HuggingPapers]

Tencent ARC Lab just released GameHorizon Suite

A unified data and evaluation suite measuring AAA gameplay capabilities across multiple temporal horizons for VLMs, UMMs, GUI, coding, and game agents.

[Diogo Almeida @CompleteSkeptic]

After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?

I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev

• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions

AFAICT the shortest path to AI-based economic revolution

[OpenAI @OpenAI]

This is GPT-6 Astra.

Anything you can do on a computer, Astra can do for you. Fast.

[@levelsio @levelsio]

Today is a very historical moment for AI video generation

You can now generate AI video faster than you can watch it

Before it'd take let's say 2-5 minutes to generate 15 seconds of video

@fal made a post-trained Minimax H3 variant called Max which is 50x faster than the original but still maintains quality

It generates 15 seconds of video in 9 seconds!

That means you can now do new things like build a perpetual livestream with it that never ends!

[Xiaodong Cun @shadocun]

Introduce #EditaLive, our new model for streaming character animation with instruction-based Editing (single forward!!!）. Arxiv: https://t.co/dU3Z28rGXV; Code: https://t.co/g2cLsxS7CX

[Google DeepMind @GoogleDeepMind]

We’re reimagining a 50-year-old interface - the mouse pointer - with AI. 🖱️

These experimental demos show how people can intuitively direct Gemini on their screens using motion, speech, and natural shorthand to get things done 🧵

[Ying Shan @yshan2u]

Pixal3D (SIGGRAPH’26): a new image-to-3D paradigm for high-fidelity 3D asset creation. Demo, code, paper available!

[Xiaodong Cun @shadocun]

#CutClaw

[DailyPapers @HuggingPapers]

CutClaw: Agentic Hours-Long Video Editing via Music Synchronization

An autonomous multi-agent framework that transforms hours of raw footage into cinematic montages. It leverages MLLMs as a Playwriter, Editor, and Reviewer to orchestrate storytelling and synchronize cuts with music beats.

[Xiaodong Cun @shadocun]

#PersonaLive

[AI Search @aisearchio]

This AI allows you to stream as anyone in realtime.

Free & open source, works with regular GPU.

It was a pain to install but I eventually got it to work. Full tutorial: https://t.co/CiKdjY8F7B

[Chen Sun 🤖 @ChenSun92]

Cute RL paper from Disney Research for the holidays❄️

"The illusion of believability is fragile: even small inconsistencies, such as rough foot impacts or jitter, can break the character's lifelike appearance."

To maximize the character's believability, they used RL to optimize for realism & acoustic stealth.

They basically conditioned the policy directly on artist-created animation files, training the agent to learn Olaf's specific rolling gait from the movie rather than a standard efficient robot walk.

To solve the "robotic clanking" problem, they introduced an "impact reduction" reward that penalizes high vertical velocity changes specifically at the moment of foot contact, effectively teaching the robot to "tiptoe".

Happy holidays / Merry Christmas!🎄❄️

[Tom Dörr @tom_doerr]

Portrait image animation for live streaming

https://t.co/DUNCjnQfMM

[Xiaodong Cun @shadocun]

🥳🥳 code is available at : https://t.co/Mi1AKTURzn

[ハカセ アイ(Ai-Hakase)🐾生成AI・ComfyUI @ai_hakase_]

PersonaLive登場！リアルタイムAI換人で配信革命！✨
via:
驚異のAI技術が爆誕！動画内の人物を別アバターにリアルタイム変換します。Vtuberやクリエイター必見ですよ！🚀 #PersonaLive #生成AI

[Wildminder @wildmindai]

ComfyUI nodes for PersonaLive: Expressive Portrait Image Animation for Live Streaming.
https://t.co/rof4ve1EMb

[Gorden Sun @Gorden_Sun]

PersonaLive：实时换人
基于LivePortrait，实时视频+参考人物照片，把视频换成参考人物，可以认为是Wan 2.2 Animate的实时版本。
Github：https://t.co/HKbONbrhc3
模型：https://t.co/zBZAdSP5Sv

[Xiaodong Cun @shadocun]

😊😊😊👏👏

[DailyPapers @HuggingPapers]

PersonaLive! Expressive Portrait Animation

A novel diffusion framework by GVC Lab & University of Macau researchers enables real-time, streaming, infinite-length portrait animations with 7-22x speedup on a single 12GB GPU.

[Xiaodong Cun @shadocun]

Welcome to try our new paper for real-time，streaming image portrait animation.😊😊😊🎊

[toyxyz @toyxyz3]

https://t.co/WZofk3rbvq

[toyxyz @toyxyz3]

https://t.co/WZofk3rbvq

[Wildminder @wildmindai]

Wan again? No, it's SD! PersonaLive!: Your vtuber dreams are now real-time, streamable & infinite-length;
- animates any portrait at 15.82 FPS with just 0.253s latency
https://t.co/qgseSefV7Z