Blog
Page 10 of 32
TTS evaluation for voice agents: beyond MOS to a phone-ready scorecard
TTS evaluation for voice agents: why MOS misleads, round-trip intelligibility, entity tests, 8 kHz phone checks, streaming TTFB and a copyable scorecard.
LiveKit vs Vapi (2026): Pricing, Latency, Lock-In, and a Fair Way to Test Both
LiveKit vs Vapi with verified 2026 pricing, TCO math at 10k, 100k and 1M minutes, per-hop latency, turn-detection internals, lock-in and a fair test plan.
The Voice Agent Stack: Every Layer, the Latency Budget and the Real Cost (2026)
Every layer of a voice agent stack, from SIP and codecs to turn detection and evals, with an 800 ms latency budget and verified per-minute costs.
LiveKit voice agent testing guide: unit tests, simulations, SIP and load (2026)
How to test LiveKit voice agents: pytest helpers, judges, audio simulations, SIP, lk load tests, and turn-detector traps that differ between dev and prod.
What to Log on Every Voice Agent Call: The Complete Schema
The voice agent call logging schema: call, turn, event and artifact fields, join IDs, OpenTelemetry code, redaction rules, and the 10 queries to run.
Pipecat Cloud vs LiveKit Cloud: Managed Voice Agent Hosting Compared
Pipecat Cloud vs LiveKit Cloud, side by side: deploy CLI, cold starts, regions, telephony, logs, compliance, pricing model, lock-in, and a pilot plan.
Can You Use LiveKit and Pipecat Together? The Hybrid Voice Agent Stack
Yes, LiveKit and Pipecat work together. A build guide for Pipecat on LiveKit transport: SIP dispatch, tokens, outbound calls, and the seams that break.
LiveKit vs Pipecat Latency: Where the Milliseconds Actually Go
LiveKit vs Pipecat latency, turn by turn: why the framework is a small slice, which knobs move p95, and how to cut delay without cutting callers off.
LiveKit vs Pipecat at Scale: Concurrency, Workers, and Production Load
LiveKit vs Pipecat scaling from 10 to 10,000 calls: agent servers vs per-session processes, cold starts, draining, rate limits, and a capacity worksheet.