Today's curated cutting-edge AI information, welcome to read 👇
🎬 ByteDance releases Doubao video generation models: PixelDance and Seaweed, supporting complex actions and multi-shot video generation.
👤 HelloMeme: New face migration model, improving image and video quality, can be integrated with SD1.5's Lora model.
📱 BBackupp: Open-source iOS device backup tool, supporting external hard drive backup and encryption, facilitating device migration.
🖥️ Compiler Explorer: Interactive programming language compiler exploration tool, supporting 30+ languages, real-time viewing of assembly code.
Cutting-edge News
1. ByteDance releases two Doubao video generation models, the demo looks impressive!
The models are PixelDance and Seaweed, with PixelDance being more advanced, both based on the DiT architecture.
They can achieve complex continuous action generation for characters, as well as multi-shot combined videos, and even extreme control over camera movements.
Here are a few highlights:
- Supports smooth transition between multiple shots in a single prompt while maintaining consistency in theme, style, and atmosphere;
- Supports one-time generation of 10s complete story videos;
- Supports generation of various video styles, such as black and white, 3D animation, 2D animation, Chinese painting, etc.;
- Supports generation of various video ratios, including 1:1, 3:4, 4:3, 16:9, 9:16, 21:9 six ratios.
For detailed introduction and demo videos, check the following links:
PixelDance: https://console.volcengine.com/ark/region:ark+cn-beijing/model/detail?Id=doubao-pixeldance
Seaweed: https://console.volcengine.com/ark/region:ark+cn-beijing/model/detail?Id=doubao-seaweed
Currently, only enterprise users are invited for testing on Volcano Engine. It will take some time before it's fully open to individual users on JiMeng.
Cutting-edge Technology
- Another effective face migration model: HelloMeme.
Fine-tuned based on the Animatediff model, it improves the quality and continuity of image and video generation, and can seamlessly integrate with Lora models based on SD1.5.
Detailed introduction: https://songkey.github.io/hellomeme/
In addition, it can generate talking videos driven by voice from the generated images, with lip-syncing, which looks quite good.
Open Source Projects
1. An open-source free tool for easy backup of iOS devices: BBackupp.
It supports backup to external hard drives, monitoring backup progress, encrypting backups for more security, and seamless migration to new devices.
GitHub: https://github.com/Lakr233/BBackupp
For those who recently switched to a new iOS device and need to backup, you can try this tool.
2. Sharing an interactive programming language compiler exploration tool: Compiler Explorer.
It allows interactive running and exploration of compilers for multiple programming languages in a web browser, supporting over 30 languages, including C, C++, Rust, Go, Python, Java, Swift, and more.
GitHub: https://github.com/compiler-explorer/compiler-explorer
Online experience: https://godbolt.org/
Moreover, it provides various editors, tools, and visualization options, allowing online code editing and real-time viewing of compiled assembly code.