Albert Xu

Film Producer & Director of Navigator Pictures.
Running to the Future.

[Albert Xu @iamAlbertXu]

I’m excited to finally share THE TIDE OF THE DRAGON — an 18-minute AI Chinese Mythic Fantasy film I’ve been working on.

This was my attempt to push AI video beyond short visual demos into real storytelling.

Watch the 18-minute Full film in the reply.👇

[Albert Xu @iamAlbertXu]

Watch the full 18-minute film here:
The Tide Of The Dragon https://t.co/ydQUCk2fDr 来自 @YouTube

[Albert Xu @iamAlbertXu]

I’m excited to finally share THE TIDE OF THE DRAGON — an 18-minute AI Chinese Mythic Fantasy film I’ve been working on.

This was my attempt to push AI video beyond short visual demos into real storytelling.

Watch the 18-minute Full film in the reply.👇

[LTX @ltx_model]

LTX HDR beta is now live.

Every AI video model before this one output 8-bit SDR only. Fine for social clips. The format falls apart the moment you try to grade. Highlights clip. Shadows crush. AI footage won't composite cleanly against higher-bit-depth CGI.

Resolution was never the real issue. Dynamic range was.

[Albert Xu @iamAlbertXu]

@freepik Feature suggestion:
-Selective cropping and saving of images generated from a 3x3 grid;
-Quick access to this function via right-click menu;
-Ability to select and save only a few images instead of all; -Improved user experience of the existing variation function.

[Albert Xu @iamAlbertXu]

@freepik Feature suggestion:
-Selective cropping and saving of images generated from a 3x3 grid;
-Quick access to this function via right-click menu;
-Ability to select and save only a few images instead of all; -Improved user experience of the existing variation function.

[Google DeepMind @GoogleDeepMind]

Step inside Project Genie: our experimental research prototype that lets you create, edit, and explore virtual worlds. 🌎

[Dinda Prasetyo @heydin_ai]

This time, I’m sharing the storyboard from my experimental AI film, Kair, and how I use FLUX.2 Pro in @Adobe @adobefirefly Boards to develop the images during the early storyboard phase.

I also use FLUX.2 Pro in @photoshop to generate 2×2 stacked storyboard grids, exploring multiple frames, lighting, and compositions in one go. It’s a fast, efficient way to iterate while keeping a consistent, cinematic look. More in the comment below 👇

#AdobeFireflyAmbassadors #Ad #AdobeFirefly #FLUX2 #AdobePartnerModels

[Xiao Fu @lemonaddie0909]

Video generation, but 4D, dynamic, scene-consistent, and very long at the same time?!

Introducing 𝐏𝐥𝐞𝐧𝐨𝐩𝐭𝐢𝐜𝐃𝐫𝐞𝐚𝐦𝐞𝐫, 𝐦𝐮𝐥𝐭𝐢-𝐯𝐢𝐞𝐰 𝐯𝐢𝐝𝐞𝐨 𝐠𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐨𝐧 𝐰𝐢𝐭𝐡 𝐥𝐨𝐧𝐠-𝐭𝐞𝐫𝐦 𝐬𝐩𝐚𝐭𝐢𝐨-𝐭𝐞𝐦𝐩𝐨𝐫𝐚𝐥 𝐦𝐞𝐦𝐨𝐫𝐲! The scaling secret is very simple: an autoregressive paradigm with minimal 3D inductive bias, aided with a spatially grounded memory retrieval mechanism.

🌐 Project page: https://t.co/2h2YQdAQez
🌐 Paper: https://t.co/aIqhokIsP7

[Higgsfield AI 🧩 @higgsfield_ai]

Meet Higgsfield Relight.
Professional lighting control without the studio.

Pick a direction, adjust intensity, set color temperature. We handle the rest: 6 quick presets, precise 3D positioning, soft-to-hard shadow control.

retweet & reply & follow & like for 220 creds

[傅盛 @FuSheng_0306]

李飞飞：K12教育完全是在浪费学生的时间，AI在快速证明很多东西是机器可以做到的，让人类花十几年甚至几十年的时间做一大半机器都能做到的事情，这是对人类的一种浪费。

[DailyPapers @HuggingPapers]

Kling Team (Kuaishou Technology) introduces MemFlow

MemFlow uses adaptive memory to dynamically retrieve relevant historical frames, ensuring consistent & efficient long streaming video narratives, even with event or scenario switches.

[DailyPapers @HuggingPapers]

InfCam: Cinematic camera control for AI video

Tired of jittery AI camera movements?

InfCam uses infinite-homography warping to generate camera-pose faithful videos without relying on depth priors.

Control rotations and translations directly in the 2D latent space for stable, high-quality results.

[Wildminder @wildmindai]

DreaMontage by ByteDance: frame-guided video generator;
- creates seamless long takes from images/clips at any timestamp;
- outperforms Vidu Q2, Pixverse V5 & Kling 2.5.
No code :(
https://t.co/EGGYo6xc3y

[AIGCLINK @aigclink]

微软开源了TRELLIS.2，图像到3D生成能力很强，一张图生成带材质的高精3D模型，速度快，可以处理复杂拓扑

4B，可以生成512³–1536³分辨率
H100上，生成512³分辨率约3s，1024³约17s，1536³约1min

输入一张图
输出可直接拖进Blender、Unity、Unreal的.glb三角形网格
附带PBR四件套贴图：颜色、粗糙度、金属度、透明度

比如说用到电商场景，就可以拍照、拖进TRELLIS.2、1分钟拿glb，然后直接上架

#图生3D #TRELLIS2

[Albert Xu @iamAlbertXu]

@D_studioproject Thanks! What's your prompt to upscale and refine a frame in the grid?

[Tengfei Wang @ ICML 🇰🇷 @DylanTFWang]

🎮Get a first look at Tencent HY World 1.5 (WorldPlay)! 🎮
Our newest world model with real-time interaction and long-term memory.

It’s going *open-source* tomorrow.

[DailyPapers @HuggingPapers]

EgoX: Generate immersive first-person video from any third-person clip

A novel framework from KAIST AI & Seoul National University that leverages video diffusion models to transform a single exocentric video into a realistic egocentric view. See it in action!

[DailyPapers @HuggingPapers]

Huawei's MoCapAnything: Unified 3D Motion Capture for Any Skeleton

Imagine animating anything from a monocular video! This new framework from Huawei tackles Category-Agnostic Motion Capture, producing high-quality skeletal animations & cross-species motion retargeting across diverse rigs.

[DailyPapers @HuggingPapers]

RealGen: Eliminating AI artifacts for truly photorealistic images

Introducing a new text-to-image framework that uses detector-guided rewards to overcome "fake" AI looks like overly smooth skin. It significantly boosts realism, detail, and aesthetics, outperforming top models.

[大雄 @dx8152]

The first LoRa code for Qwen-Edit-2509 light migration on the entire internet, solving the problem of relighting the model. This is a very powerful tool, removing the artistic lighting from the original lighting migration reference image.
@Ali_TongyiLab