Reduce latency and packet loss in audio streams when dealing with jitter over unreliable networks.FreeSWITCH's improved elastic jitter buffer system.
OpenSIPS 4.0 introduces some of the most significant architectural changes in recent years. This session takes a deep technical look at the new internals and operational capabilities now available to SIP platform engineers. Topics include the redesigned TCP/TLS subsystem, realtime profiling and observability tooling, PRACK/UPDATE interworking, bond sockets, Proxy Protocol support and improvements targeting modern distributed SIP deployments.
This session explores how to safely run real-time Voice AI inside deterministic telephony systems. We present an open-source production-oriented architecture built on: a SIP edge, FreeSWITCH and a streaming AI pipeline. FreeSWITCH serves as the deterministic media and call-control layer, enforcing routing rules, timeouts, fallback paths, and session state. Live audio is streamed into an AI pipeline for STT → inference/translation → TTS, with streaming responses injected back into the call. We will cover: - Bridging RTP audio into a streaming AI pipeline - Defining conversational SLOs - Instrumenting stage-level metrics and visualizing them - Isolating AI failures and enforcing deterministic fallbacks in real-time telephony. The session concludes with a live demo of a real-time voice solution running on FreeSWITCH, complete with signaling visibility, media metrics, and AI pipeline timing. This talk is aimed at engineers building measurable, resilient Voice AI systems on open telephony infrastructure.