{"type":"rich","version":"1.0","provider_name":"Transistor","provider_url":"https://transistor.fm","author_name":"Tech Stories Tech Brief By HackerNoon","title":"How to Evaluate STT for Voice Agents in Production","html":"","width":"100%","height":180,"duration":837,"description":"\n This story was originally published on HackerNoon at: https://hackernoon.com/how-to-evaluate-stt-for-voice-agents-in-production.\nMost STT benchmarks measure the wrong thing. Here's how to evaluate speech-to-text for voice agents using the metrics that actually drive production performance\nCheck more stories related to tech-stories at: https://hackernoon.com/c/tech-stories.\n You can also check exclusive content about #ai-voice-agent, #voice-agent-stt, #pipecat, #voice-ai, #conversational-ai, #ai-voice-agent-benchmarking, #stt-evaluation-metrics, #good-company, and more.\nThis story was written by: @speechmatics. Learn more about this writer by checking @speechmatics's about page,\n and for more stories, please visit hackernoon.com.\nVoice agent developers are optimising for TTFB — time to first byte — but it's one of the least useful metrics in production. \n\nWhat actually determines how fast and reliable your agent feels is TTFS (time to final segment): the gap between a user finishing speech and a stable transcript landing in your LLM. \n\nThis piece breaks down the Pipecat benchmark — currently the most credible public eval for STT in voice agents — explains semantic WER and why it beats standard word error rate for this use case, and makes the case that accuracy and latency are inseparable. A faster wrong answer is still a wrong answer.\n \n ","thumbnail_url":"https://img.transistorcdn.com/IuqXIpaNNuezY7jNfIDnL5gqB1iL_SEndwUUzLGdljY/rs:fill:0:0:1/w:400/h:400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS9zaG93/LzQxNDI5LzE2ODM1/ODM0NjQtYXJ0d29y/ay5qcGc.webp","thumbnail_width":300,"thumbnail_height":300}