Livekit

Conversational intelligence split into Audio turn detection (EOT) at $0.50 per 1,000 requests and Adaptive Interruption at $1.20 per 1,000 requests for self-hosted agents; free on LiveKit Cloud.

New Twilio connector and WhatsApp connector add-ons let customers bridge Twilio Programmable Voice and WhatsApp Business calls into LiveKit rooms, tiered by concurrent calls (5/20/100/Custom) with $0.004/min and $0.003/min overage pricing.

Scale plan's LiveKit Inference concurrency limit (50 concurrent sessions) now notes customers can request more via the dashboard, matching existing language on the Concurrent agent sessions row

Added two LiveKit Inference features: Custom voices (record/upload custom TTS voices - 20 included on Ship, 50 on Scale) and Zero data retention (prompts, audio, model outputs not logged or stored).

Quarterly roll-up · includes Jun 1 weekly, Apr 13 weekly

Audio feature descriptions updated with specific provider/model specs: Background noise suppression now specifies Krisp NC and ai-coustics QUAIL_L; Voice isolation specifies Krisp BVC and ai-coustics Voice Focus 2.1; Enhanced noise cancellation updated from 'Krisp' to 'Krisp NC'.

Two agent audio features renamed: 'Audio enhancement' is now 'Background noise suppression'; 'Speaker isolation' is now 'Voice isolation', dropping the ai-coustics Voice Focus model reference.

Concurrent agent sessions on Scale plan reduced from Up to 1,000 to Up to 600. Enterprise agent deployments and concurrent connections changed from Unlimited to Custom. Audio enhancement, Speaker isolation, and Conversational intelligence features added for agents. Inference discounts added to Scale plan. Agent observability BETA free period removed.

Quarterly roll-up · includes Mar 23 weekly, Feb 16 weekly, Jan 5 weekly

Added audio enhancement features including Speaker isolation using ai-coustics Voice Focus to suppress background voices [Speaker isolation: 100 mins (Ship), 1,000 mins (Scale), 10,000 mins (Enterprise) at $0.0012/min overage; Audio enhancement improves STT accuracy and VAD precision; Conversational intelligence enables end-of-turn detection and interruption handling]

Rebranded Inference to LiveKit Inference; added Inference discounts to Scale plan [Renamed section headers and feature names; Scale plan gained 'Inference discounts' feature; Enterprise 'Volume pricing' clarified to include inference]

Agent Observability exited beta: free-until-Jan-2026 notice removed (overages now billed); Observability row added to pricing calculator as a toggleable line item.

Telephony section restructured - SIP participant minutes replaced with dedicated US local and toll-free phone number pricing Agent observability feature added as new capability (Beta, free until January 1, 2026)

Primary usage metric changed from 'connection minutes' (time connected to network) to 'agent session minutes' (time in session with AI agent) LiveKit restructured pricing from connection-based model (concurrent participants, connection minutes, bandwidth) to AI agent-centric model (agent session minutes, inference credits, agent deployments)

DPA (Data Processing Addendum) availability explicitly shown for all plans - 'Standard' DPA for Build, Ship, and Scale plans; 'Custom' DPA for Enterprise

Noise cancellation feature renamed from 'Noise cancellation with Krisp' to 'Enhanced noise cancellation' (Krisp branding removed) Concurrent imports limit added: Build plan gets 2, Ship plan gets 100, Scale and Enterprise get Unlimited

Stream import transcode time (8,000 minutes with video/audio overage pricing) added to Scale plan Stream import transcode time (600 minutes with video/audio overage pricing) added to Ship plan