no you ng

[Hugging Models @HuggingModels]

Meet a game-changer for portrait artists and photographers. This AI model transforms ordinary photos into ultra-realistic portraits with stunning detail. It's not just another filter, it's a professional-grade image-to-image pipeline that understands human features.

[Sket_h @SketchsArt]

A combination of 2 addons that I plan to release soon.
1- Real time Camera tracking via iphone
2-Smart DSLR like autofocus for Blender Camera
Stay tuned!
#blender #cgi #b3d #3danimation #animation

[Spencer Sterling @cerspense]

This is what directing an AI creative studio looks like.

I gave my orchestration system a reference image and had it deploy three different interpretations across three machines running Blender simultaneously, all visible from a single dashboard.

When one direction looks promising, I branch from it. The system copies the files to the other machines and keeps exploring from that new starting point. Branch, direct, branch again. Each split opens up a new creative path while I give notes in real-time.

The number of creative directions I can explore simultaneously is only limited by how many computers I have.

[World Labs @theworldlabs]

We’re building foundational world models to power the next era of 3D. From robotics to gaming, spatial intelligence unlocks entirely new worlds.

Powered by inference at scale – shoutout to Baseten.

[Keenan Crane @keenanisalive]

Fun & likely unintended use case for #NanoBananaPro: image→3D point cloud

Prompt: This is a rendering of a 3D model of a person. Make an accurate image of the corresponding normal map.

Then I run Poisson-based normals→3D

Pretty amazing for a general-purpose 2D image model 🤩

[Wildminder @wildmindai]

Code2Worlds. language-to-simulation framework.
- text to simulation scripts for Blender using Gemini 3;
- better scene richness; slashes physics failure;
- uses VLM-Motion critic.
https://t.co/vcjqfWRs0Z

[Hi3D @Hitem3D]

🚀Hitem3D Portrait 2.1 Model Update！

This update focuses on improving realism in human portrait generation.

✅Key Improvements：

👉🏻Improved accuracy of facial features
👉🏻More natural and balanced facial proportions
👉🏻More realistic body posture and overall structure

🚀try it now：https://t.co/4dFy7Vxy46

#hitem3d #3DPrinting #3dmodeling #ai3d

[KIRI Engine - 3D Scanner App @KIRI_Engine_App]

Stylizing Gaussian Splats:
Rather than tricky style transfer techniques or Gen AI, you can use simple image filters when training 3DGS 🎨

The trick is to create the sparse points with the featureful original images - then swap out for the filtered pics during training💡

[傅盛 @FuSheng_0306]

Manus联合创始人季毅超：很多创业者“没有乔布斯的命，却得了乔布斯的病”，但AI时代创始人最稀缺的特质是“正常”。移动互联网时代偏爱艺术家，靠低边际成本赌一把，AI 创业更像传统制造业，面临极高的算力成本和经营压力。在这种环境下，相信常识、身心健全、能冷静面对受挫并迅速站起来的“正常人”，反而拥有更强的韧性，能带着团队走得更远。

[迈克 Mike Chong @mike_chong_zh]

LinkedIn 上有人分享了如何构建 99% 代码都是 AI 来写的系统。系统性思维很棒，翻译了一下：

到现在我已经用 Claude Code 和 Copilot 来写代码差不多一个月了，基本就是我自己几乎不写——99.9% 的代码都是 AI agent 写的。下面是我能靠 AI 干成的事情，以及大概的花费：

Claude Code：大约 12.5 亿 Tokens，花了 ~640 美元
Copilot：1375 次高级请求，企业版（39 美元/月）
Claude Chat：几百次对话，同样是企业版（应该和 Copilot 一样）

代码库的变化：
* 新增 197K 行代码
* 删除 208K 行代码
* 主分支上有 355 次提交
* 合并了 180 个 PR
* 解决了 98 个 GitHub issue

我是怎么做到的？

1) 所有代码都必须过一套非常严格的 CI/CD 流程：静态检查、格式化、代码安全工具，还有完整的测试集
2) 代码库已经模块化了，每个组件都通过清晰的接口来调用
3) 接口和实现完全分开
4) 每个组件都有自己的文档
5) 每个组件的测试、文档、头文件、实现都放在一起
6) 每个组件保持在 5k 行代码以下（包括文档）
7) 每个 GitHub issue 都写得很详细，带代码引用和问题描述
8) 搭建了一个 7 层的提示体系，记录各种���范：工具层（Copilot vs Claude）、语言层（C++ vs Python）、项目层（按仓库）、角色层（开发 vs 测试 vs 架构 vs 审查）、组件层（按目录）、任务层（对应 GitHub issue）、执行层（具体操作）

这样，我只需要分配任务，15-30 分钟就能拿到第一版，然后我再迭代、审查、验证，最后一小时内就能提交——几乎不用我亲手敲代码。

现在真正的瓶颈是我自己，因为我审查代码的速度跟不上。

而且 Claude Code 现在已经包含在企业计划里了。是时候继续玩一玩，让我的“Agent 军团”来帮我缓解这个瓶颈。

[Common Sense Machines @CSM_ai]

GPT-4o Image Generation to Part-based 3D Characters with PBR, in under 10 minutes ⚡️

Workflow:
🎨 Prompt GPT-4o to get an image (e.g., "3D asset of a styled character with all parts laid on a sheet for image to 3D")
🧩 Use CSM AI's Part-based tool to generate parts and assemble in Blender.

[Justine Moore @venturetwins]

🚨 New @a16z thesis: AI avatars

Generative models are finally good enough to make talking characters with AI.

Combining a voice with a face has unlocked countless use cases - from sales reps for enterprises to consumer animation.

Our market map + insights 👇

[Rohit Gandikota @rohitgandikota]

Can you ask a Diffusion Model to break down a concept? 👀

SliderSpace 🚀 reveals maps of the visual knowledge naturally encoded within diffusion models. It works by decomposing the model's capabilities into intuitive, composable sliders.

Here's how 🧵👇

[Nodes Interactive @nds_interactive]

Hair Press Collider done with #geometrynodes. Grab it for free here: https://t.co/bgixwBVS4n
#blender3d #b3d #3d #3dart #3dmodeling #3dartwork #indiegame #indiegamedev #3dartist #gamedev #unrealengine #unity #geonodes #procedural #3danimation

[Poly Hammer @PolyHammerTools]

🚀 Blender Addon for Metahumans!

Get ready to fully customize your Metahumans in Blender! Our new addon allows you to import DNA files and offers a seamless 1-to-1 integration with the face board rig logic, just like in @UnrealEngine. Now, you can work in Blender and effortlessly push changes back to your Unreal project in 1-click!

🎉 Early Access Opportunity! We are wrapping up development, but you can receive this addon for FREE upon release by simply helping us share this post!

👉 Sign up within the next 30 days to receive the addon FREE!
https://t.co/wgH6bl9v2G

🔔 Follow us! Stay Connected for More Demos and Tutorials!

#DigitalHumans #3DAnimation #b3d #blender3d #3d #gamedev #gamedevelopment #blenderartist #UnrealEngine #UE5

[Gizem Akdag @gizakdag]

Now she really has a beautiful smile.

Midjourney + Magnific + Hailuo

I also added the before/after Magnific images to the thread.

[硅谷王川 Chuan @Svwang1]

人需要把所有急功近利赚快钱的方式全部尝试几遍，并不断最后亏钱，亏得心服口服以后，才能被迫选择到短期看似缓慢，但长期并不慢的正道上。这个过程至少需要 15-20年，否则总还是有贪多求快的侥幸心理。

所以 40岁以下的人，很难真正实践稳健的长期投资。即使嘴上认同，实操上根本按捺不住短线操作快感幻觉的诱惑。

[Shalini Tewari @maybeshalinii]

HOW TO STRUCTURE YOUR BACKEND CODE IN NODE.JS(express.js).
[with explanation]

📁
├── 📄 app.js
├── 📁 bin
├── 📁 config
├── 📁 controllers
│ ├── 📄 customer.js
│ ├── 📄 product.js
│ └── ...
├── 📁 middleware
│ ├── 📄 auth.js
│ ├── 📄 logger.js
│ └── ...
├── 📁 models
│ ├── 📄 customer.js
│ ├── 📄 Product.js
│ └── ...
├── 📁 routes
│ ├── 📄 api.js
│ ├── 📄 auth.js
│ └── ...
├── 📁 public
│ ├── 📁 css
│ ├── 📁 js
│ ├── 📁 images
│ └── ...
├── 📁 views
│ ├── 📄 index.ejs
│ ├── 📄 product.ejs
│ └── ...
├── 📁 tests
│ ├── 📁 unit
│ ├── 📁 integration
│ ├── 📁 e2e
│ └── ...
├── 📁 utils
│ ├── 📄 validation.js
│ ├── 📄 helpers.js
│ └── ...
└── 📁 node_modules

Here's a brief explanation:
app.js
It's like the control centre of your web application. It's where you set up and manage everything.

bin
Think of this as a place for starting your web server. It's where you have scripts that make your website work.

config
These are like settings for your website, such as where your database is or how your website should behave.

controllers
This is where you put the brains of your website explains the name controllers. Each file here handles a different part of your site, like customer stuff or product stuff.

middleware
Imagine these as helpers that help your website do things like checking if you're logged in or keeping a record of what people do on your site.

models
This is where you describe what your data looks like. If your website is a store, this is where you say what a product is or what a customer is.

routes
These are like the paths to different parts of the website. If the website is a city, these are the streets and highways.

public
Think of this as your storage room for things everyone can see, like images, styles, and scripts.

views
If your website is like a book, these are the pages. This is where you put together what people see on the screen.

tests
These are like exams for your website to make sure it works correctly. You create different kinds of tests to check different parts of your site.

utils
These are like handy tools you use to make your website better, like checking if someone's email is valid or formatting dates nicely.

node_modules
This is like your toolbox filled with tools (libraries and code) that you use to build your website. It's automatically filled with things your project needs.

This organized structure helps you build and manage your website in a neat and organised way, making it easier to understand and work on, especially when your project gets bigger and more complex and you have to make changes very often.
I am not saying that this is the only or the best way, this is what I usually follow and thought about sharing it.

[Jiayuan (JY) Zhang @jiayuan_jy]

如何更高效地阅读论文 📃

上周 @CopilotHubAI 新发布的 Magic Copilot 功能可能会永远改变你阅读论文的方式，只需要 3 个步骤，就可以开始交互式地阅读论文：

1. 上传论文 PDF
2. Magic Copilot 自动生成 Prompt
3. 开始阅读

这个 thread 以 ChatDB 这���论文为例 👇
https://t.co/AezvoY9fBi

[歸藏(guizang.ai) @op7418]

Meta AI 相当高产啊，开源了 DINOv2 视觉模型和FACET（计算机视觉公平性评估）测试。

DINOv2 是一种计算机视觉模型，通过自我监督学习进行训练，可产生通用特征。发布了一系列基于 DINOv2 的密集预测模型，用于语义图像分割和单目深度估计。

FACET（计算机视觉评估公平性），这是一个新的综合基准，用于评估计算机视觉模型在分类、检测、实例分割和视觉基础任务方面的公平性。该数据集由包含 50,000 人的 32,000 张图像组成。

DINOv2 论文：https://t.co/UrGHanWprM
DINOv2 代码：https://t.co/cpAxLvHiN0
DINOv2 演示：https://t.co/gJQgstDdrO
FACET 论文：https://t.co/wCkq44l6XN
获取 FACET 数据集：https://t.co/ltdKzE9XVX