The speaker, Svan, discusses Almond Call, a video calling app developed by Almond. He explains that Almond Call uses the RTC for media transmission and matrix for action signaling. Although the app has many desirable features, such as screen sharing and encryption, it faces drawbacks due to the full mesh connection model it uses, which requires a lot of upstream bandwidth. The speaker proposes using a backend called the focus, which is either an MCU or an SFU. They explain that an SFU is a better solution as it is more reliable, requires less upstream bandwidth, and allows clients to have control over their streams. The speaker introduces Waterfall, an SFU contributed by Sean, the creator of Pion. Waterfall uses the matrix protocol for signaling and supports cascading, allowing SFUs to connect to other SFUs and create federated matrix RTC worlds. The speaker also discusses how clients connect to a focus, choose an SFU, and establish connections using device messages over matrix and WebRTC data channels. They explain how the SFU forwards the ideal resolution of a stream based on client preferences and available bandwidth. The speaker concludes by mentioning future plans, such as focus selection logic and cascading, implementing end-to-end encryption via insertable streams, and improving track switching. They also invite the audience to contribute to Element Call by checking out the stable and development versions, exploring repositories and pull requests, and sharing their expertise. The presentation concludes with a live demo of Element Call, showcasing features like freedom mode, spotlight mode, and screen sharing.