puck VS co-tracker

See the full changelog for all changes via the GitHub release.

If you’re starting from scratch, you can also use one of the Puck recipes to spin up a new project:

⭐ Support us on GitHub by dropping a star

Puck is an embeddable, modular, open-source visual editor for React with built-in support for complex drag-and-drop page building. While GrapesJS provides a barebones page building experience, and tools like Builder.io and Storyblok offer fully-fledged CMS platforms, Puck aims to bridge the gap. It combines an extendable ready to use page editor with a fully decoupled page export model—giving you flexibility without locking you into a specific backend or proprietary ecosystem.

That would be wonderful, but will require some further work. Puck doesn't provide a grid, but it supports user grid implementations.

To support something like you're suggesting would likely require an official Grid component, which I'm now tracking here: https://github.com/measuredco/puck/issues/843

The official next recipe now uses Next 15.1 and React 19 by default. If you're using the recipe and want to upgrade, make sure you also upgrade to React 19 as required by Next 15 when using the App Router. See the official Next.js upgrade guide for more info.

I was using another great library called puck, which is basically a very customisable editor, that can create sites/newsletters/pdfs (something visual) and came to the next step which was turning the output of the editor into a PDF.

We're grateful for the community's support and contributions. Join the conversation on GitHub and Discord.

Hello fellow hackers!

We launched Puck in September [on Hackernews](https://news.ycombinator.com/item?id=37391848) after building it for our clients, and had a wild ride to the front page!

I wanted to share a little update on what's happened since:

1. 3.7k starts [on GitHub](https://github.com/measuredco/puck)!

🌐 Flink or Quix or RisingWave or Spark or Bytewax or ... 🌐 Important new API for Moderating by Mistral 📎 open coder 💻 Sentence Transformers Speedup 🦾 Convert an entire HTML Website to Markdown with Great GO CLI Tool 🫶 LiteLLM 📝 How to setup InstructLab Granite Model 🌐 Facebook Co-Tracker 🌐 Perplexica Search+RAG LLM 🌐 Facebook Project to watermark anything! 🌐 AI Travel Agent with LangGraph 🌐 Go + Python + AI 🌐 Graphrag Explaing 🌐 Matryoshka Embedding Detail at Multiple Scales 🌐 Browser Use for LLM 📎 Creating Advanced AI Agents with Ollama and Langchain 💻 Great command line REST Client 🤖 Inkeep Builds an AI Assistant with Milvus ✅ Visual Data Transformation ETL 💻 What is a LongRAG 🛠️ PymuPDF4llm is Your New Best Friend 🫶 WebScraping for LLM

Many point tracking models, like PIPs and PIPs++, track points independently; however, points in a video often exhibit strong dependencies, such as belonging to the same object. The original CoTracker model uses these dependencies by performing joint tracking of many points.

!git clone https://github.com/facebookresearch/co-tracker/ from cotracker.utils.visualizer import Visualizer vis = Visualizer(save_dir='.', pad_value=100) vis.visualize( video=video_tensor, tracks=pred_tracks, visibility=pred_visibility, filename='output' ); show_video("output.mp4")

📊 LLAMA 3S Instruction v0.2 🔋 Torchtune / Finetuning Models 💻 Facebook Research Co-Tracker Vision 📝 LLama 3S Instruction v0.2 🕶️ OmniParse MS 🍔 Molmo 7B-D-0924 open vision-language models 📝 IBM Granite Models on Ollama 💻 Mochi 1 Video Generation 🕶️ Text to Video Open Model 🔋 QWEN 2.5 Math 📊 QWEN 2.5 - 1.5B Instruct 🐍 Janus + Tools 🍔 MIT Han Lab VILA-U: a Unified Foundation Model Integrating Visual Understanding and Generation 7B 256 🛠️ Pixtral Multimodal model 🎃 Steiner 32B interesting model 🐦 Text to Video 👽 NVIDIA customized LLama 3.1 Nemotron 70b 👻 [Stable Diffusion 3.5 Large]https://huggingface.co/stabilityai/stable-diffusion-3.5-large) 🔋 MedEmbed for medical 🍔 Extract image PHI-3 Vision 128K

Of course Facebook has had a video tracking ML model for a year or so - Co-tracker [1] - just tracking pixels rather than segments.

[1] https://co-tracker.github.io/

Yes, I think this is very tractable for a first project. I've played around with using AI to do optical-only with pose detection models -- if I had to do it again, I would probably start with this model and try to get it running locally:

https://github.com/facebookresearch/co-tracker

This sounds like a perfect place for you to get started!

(arXiv) (GitHub)

LICENSE

Attribution-NonCommercial 4.0 International

https://github.com/facebookresearch/co-tracker/blob/main/LIC...