December 2023 Summaries

VCS is a Video Component System, a developer toolkit that enables building dynamic video compositions and multi-participant live streams. It provides an open-source web renderer package for rendering VCS directly in web applications. The system has gained dozens of features based on customer requests and ideas. A four-part series of posts will cover new features in VCS, including two new overlay graphics options, a new "highlight lines" data source, and ways to use content behind video elements. The first post focuses on motion graphics and background content, introducing the Banner and Sidebar components with various configuration options. New features include the ability to display emojis as images, control composition backgrounds using custom images or WebFrame embedded browsers, and real-time rendering of web pages. The system also supports data sources for dynamic visuals, such as highlight lines, and allows for customization of video elements and their layout. The next posts in the series will cover video layer enhancements and room-level data access.

The Zoom Video SDK, designed for developers to build browser-based video applications, has significant feature gaps and performance issues compared to more established developer SDKs like WebRTC. The SDK's proprietary approach to video encoding and decoding using custom WebAssembly modules results in high CPU usage and poor video quality, especially on older devices or variable network conditions. Additionally, the SDK lacks support for standard HTML elements, styling, and positioning of video tiles, as well as features like end-to-end encryption, HLS live streaming, and access to raw media tracks. The Zoom Web SDK's maturity is lower compared to other platforms, and its developer tooling is incomplete, with limited guides and official sample applications that require workarounds for proper functionality on Safari.

Talk to Santa Cat` is a holiday-themed mobile app that allows children to engage in live, voice-driven conversations with an AI-powered animated cat in Santa's workshop, featuring advanced Large Language Model (LLM) technologies for uniquely engaging interactions. The app was designed by Daily engineers who are also parents, aiming to create a safe and family-friendly experience, with fine-tuned AI components to provide an imaginative and age-appropriate space for children to immerse themselves in the holiday season. The app is designed with high levels of privacy, does not require login information or recording video or audio, and contains no ad monetization or tracking. To build such experiences, developers need to consider various technological challenges, including speed, accuracy, and natural conversational elements, and Daily aims to make it easier for them by providing a new open-source framework that handles the hard parts in a highly configurable way.