SIPREC Recording Server
Stream Conductor
Every conversation, captured at scale and streamed to your AI.
A lightweight, high-performance session recording server built on the IETF SIPREC standards (RFC 7865 / RFC 7866). Stream Conductor captures calls from any standards-compliant recording client and writes them to multi-channel audio, with optional vCon generation for analytics, AI and compliance. Real-time media streaming for AI analysis engines is coming next.
Why Stream Conductor
Built for Scale
An asynchronous, multi-threaded C++ core with io_uring-based file writing on Linux. Set limits on concurrent calls and streams, and use drain mode to upgrade with no dropped recordings.
Standards-Based, Not Proprietary
Works with any SIPREC client: SBCs, PBXs, contact center platforms and UCaaS services. There's no proprietary forking protocol, so you're never locked in.
AI-Ready When You Want It
Real-time media streaming API (in development): stream live, per-participant call audio straight into your own AI application while the call is still in progress.
Prefer post-call analytics? Turn on optional vCon (IETF vCon) integration and each recording becomes a structured conversation record in vConductor. Don't need either? Just record.
Technical Features
Carrier-grade SIPREC recording today, with real-time media streaming for AI coming next.
SIPREC and SIP Signaling
- IETF SIPREC session recording server (RFC 7865 / RFC 7866)
- SIP transports: UDP, TCP, TLS, WebSocket (WS) and Secure WebSocket (WSS)
- TLS 1.2 and TLS 1.3 for SIP over TLS and Secure WebSocket (WSS) (TLS 1.0/1.1 can be enabled for legacy interoperability on request)
- Multiple outbound registrations across multiple domains and proxies
- SIP Session Timers, which detect and clean up stale calls
- IPv6 support for SIP signaling
- Built on the proven reSIProcate SIP stack
Media Security
- SDES-SRTP: AES-128/192/256 with HMAC-SHA1-32/80, and AEAD AES-128/256-GCM
- DTLS-SRTP: AES-128 with HMAC-SHA1-32/80, and AEAD AES-128/256-GCM
- Optional-SRTP negotiation (RFC 4568), for wider interoperability
- TLS with ECDH, and configurable DH parameters
Media and Codecs
- G.711 µ-law/A-law (PCMU/PCMA), G.729, Opus, Speex, L16 and iLBC
- RTCP, with rtcp-mux negotiation
- ICE-lite, for interoperability with WebRTC endpoints
- RFC 4733 DTMF (telephone-event) detection, logged with timestamps and recording offsets
Recording Output
- Multi-channel WAV files, with each participant on their own channel (ideal for transcription and speaker analytics)
- SIPREC metadata saved next to each recording, including updates from re-INVITE and UPDATE
- Optional full SIP message capture, for troubleshooting client interoperability
Optional vCon and AI Integration
- Native integration with SIP Spectrum's vConductor platform
- Use vConductor's AI features to transcribe, summarize and analyze every conversation
- Optional IETF vCon creation for every recorded call, with no separate ingestion pipeline
- Recordings and SIPREC metadata delivered as structured conversation records
- High availability, with automatic failover across vConductor instances
Contact Center Integrations
- Works with any platform that supports SIPREC
- Five9 VoiceStream (work in progress): joins the agent and customer audio legs into a single recording and a single vCon
WebRTC SIPREC Recorder Plugin
- A JavaScript library that turns any WebRTC softphone into a SIPREC client
- Forks both near-end and far-end audio to Stream Conductor as separate channels, over DTLS-SRTP/ICE
- Includes a sample JsSIP softphone, so you can be up and running quickly
Operations and Monitoring
- Active Calls web page, for live diagnostics
- Prometheus metrics endpoint, ready for Grafana dashboards: sessions and streams, vCon requests, SIP traffic and retransmissions, registration health, and process CPU, memory and file descriptors
- Drain mode, for controlled, zero-loss upgrades
- Configurable limits on concurrent calls, streams and media port ranges
- Rotating log files, plus optional SIP message logging
Deployment
- Runs on Linux (a Debian .deb package is available) and on Windows
- A single configuration file, and a small footprint
How It Works
From the SIP network to multi-channel recordings and AI-ready data, in five steps.
1. Your platform forks the call
Your SBC, PBX, contact center or WebRTC softphone acts as the Session Recording Client (SRC). It opens a SIPREC session to Stream Conductor and forks each party's audio.
2. Stream Conductor records every party
It negotiates the media, including SRTP/DTLS, and records each participant to their own channel of a multi-channel WAV file.
3. Metadata is captured
SIPREC metadata (participants, streams and any updates during the call) and DTMF events are saved alongside the recording.
4. Optional: vConductor adds AI insight
If vCon is turned on, vConductor turns the call into a vCon record, and the recording is filed under its final name. vConductor's AI features can then transcribe, summarize and analyze the conversation.
5. Operations stay in control
Teams watch everything live, through the Active Calls page and Prometheus/Grafana.
Specifications at a Glance
| Standards | IETF SIPREC (RFC 7865, RFC 7866), IETF vCon, RFC 4568, RFC 5764, RFC 4733 |
| SIP transports | UDP, TCP, TLS, WS, WSS |
| TLS versions | TLS 1.2, TLS 1.3 (SIP over TLS and WSS) TLS 1.0/1.1 can be enabled for legacy interoperability on request |
| Media security | SDES-SRTP, DTLS-SRTP (incl. AES-GCM) |
| Codecs | G.711 µ/A, G.729, Opus, Speex, L16, iLBC |
| Recording format | Multi-channel WAV, plus SIPREC metadata |
| Integrations | vConductor (optional vCon), WebRTC (JsSIP) |
| Monitoring | Prometheus / Grafana, Active Calls web page |
| Platforms | Linux (.deb), Windows |
What's Next
Real-Time Media Streaming for AI (coming soon)
Stream live call audio straight to your own real-time AI engines while the call is still in progress, for live transcription, agent assist, sentiment, compliance alerts and more. Developers building real-time analysis get clean, per-participant media streams without having to touch SIP.
Video (planned)
Stream Conductor is designed to grow beyond voice. Video capture and streaming are on the roadmap.
Building a real-time AI engine? Talk to us about early access.
Stream Conductor + vConductor
Stream Conductor captures and delivers the conversation, and vConductor turns it into governed, AI-ready business intelligence.
Together they form a complete, standards-based pipeline from the SIP network to transcription, summarization, sentiment and compliance, built on IETF SIPREC and vCon with no vendor lock-in.
Stream Conductor also works fully on its own: vCon and vConductor are optional.
Sign up to stay in the loop about the latest updates