AssemblyAI STT 3.7 Realtime Agent vs ElevenLabs Scribe v3 Live

Compare

Accuracy and speed

Left gets fewer words wrong. Lower has the final text ready sooner.

Accurate and fast0ms200ms400ms600ms800ms0%2%4%6%Words wrong on clean audio (Pipecat WER)
  • STT 3.7 Realtime Agent1.68% · 47ms
  • Scribe v3 Live1.81% · 57ms

AssemblyAI STT 3.7 Realtime Agent is ahead on time to first text and final-text delay, and ElevenLabs Scribe v3 Live on both datasets together and the Ocular Dataset. The Pipecat Dataset is too close to call.

Last updated Methodology by Dileep Chagam

Who is better at what

STT 3.7 Realtime AgentScribe v3 Live
  • Overall WER2.41%1.89%
  • Pipecat Dataset WER1.68%1.81%
  • Ocular Dataset WER3.81%2.05%
  • Time to first text493ms685ms
  • Final-text delay47ms57ms

Lower is better on every measure. Too close to call means the likely ranges overlap.

Clip by clip

On every clip both models were scored on, the lower WER wins, and WERs within 5 points of each other are a draw; in the rating, a wider win counts for more. Expected win rate is what the two ELO ratings predict for a match.

STT 3.7 Realtime AgentScribe v3 Live
  • ELO rating10241001–1039 · rank 1–1610421023–1051 · rank 1–10
  • Expected win rate47%53%
  • Ocular turns WER · 168 draws of 19810 won20 won
  • Pipecat WER · 904 draws of 99949 won46 won
  • Semantic WER · 190 draws of 1983 won5 won
  • Phone line · 168 draws of 19811 won19 won
  • Call-center noise · 140 draws of 19518 won37 won
  • Babble noise · 128 draws of 19721 won48 won
  • TV noise · 144 draws of 19726 won27 won
  • Reverb · 165 draws of 1989 won24 won
  • Short replies · 61 draws of 705 won4 won
  • Entities · 48 draws of 501 won1 won

Short replies are decided on the error count and entities on exact capture, with no draw margin. How the rating is worked out

STT 3.7 Realtime Agent vs Scribe v3 Live FAQ

  • ElevenLabs Scribe v3 Live is more accurate on the public clips. On the 1,000 public Pipecat clips, AssemblyAI STT 3.7 Realtime Agent scores 2.41% and ElevenLabs Scribe v3 Live 1.89%, so ElevenLabs Scribe v3 Live makes fewer errors by 0.52 points. On the 300 licensed Ocular turns, AssemblyAI STT 3.7 Realtime Agent scores 1.68% and ElevenLabs Scribe v3 Live 1.81%, with likely ranges of 1.47 to 1.90% and 1.59 to 2.04%. The ranges overlap, so the difference is too close to call. Word error rate counts incorrect, missing, and extra words, so lower is better.

  • Yes. AssemblyAI STT 3.7 Realtime Agent finishes the transcript 192ms sooner. AssemblyAI STT 3.7 Realtime Agent has the final transcript 493ms after the speaker stops and ElevenLabs Scribe v3 Live 685ms, at the median. Both are measured on the same 206 streamed turns.

  • AssemblyAI STT 3.7 Realtime Agent finishes the transcript 192ms sooner, and the two are too close to call on real conversations. A voice agent cannot reply until the final transcript arrives, and the Ocular recordings are real conversations.

  • Yes, both return partial text before the speaker stops. AssemblyAI STT 3.7 Realtime Agent shows first text after 3.81% and ElevenLabs Scribe v3 Live after 2.05%, at the median, so ElevenLabs Scribe v3 Live is 1.77 points lower.

Test your voice agent the same way.

Cekura runs these simulated calls on your agent and scores every one.