Left gets fewer words wrong. Lower has the final text ready sooner.
Speechmatics Linden is ahead on both datasets together and time to first text, and Soniox STT-RT v5 on final-text delay. The Pipecat Dataset and The Ocular Dataset are too close to call.
Last updated Methodology by Dileep Chagam
Lower is better on every measure. Too close to call means the likely ranges overlap. Our guide explains how word error rate is measured.
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.
Short replies are decided on the error count and entities on exact capture, with no draw margin. How the rating is worked out
Speechmatics Linden is more accurate on the public clips. On the 1,000 public Pipecat clips, Soniox STT-RT v5 scores 3.02% and Speechmatics Linden 2.97%, so Speechmatics Linden makes fewer errors by 0.06 points. On the 300 licensed Ocular turns, Soniox STT-RT v5 scores 2.54% and Speechmatics Linden 2.46%, with likely ranges of 2.27 to 2.83% and 2.19 to 2.73%. The ranges overlap, so the difference is too close to call. Word error rate counts incorrect, missing, and extra words, so lower is better.
Speechmatics Linden is more accurate on the public clips. On the 1,000 public Pipecat clips, Soniox STT-RT v5 scores 3.02% and Speechmatics Linden 2.97%, so Speechmatics Linden makes fewer errors by 0.06 points. On the 300 licensed Ocular turns, Soniox STT-RT v5 scores 2.54% and Speechmatics Linden 2.46%, with likely ranges of 2.27 to 2.83% and 2.19 to 2.73%. The ranges overlap, so the difference is too close to call. Word error rate counts incorrect, missing, and extra words, so lower is better.
No. Speechmatics Linden finishes the transcript 5ms sooner. Soniox STT-RT v5 has the final transcript 1.20s after the speaker stops and Speechmatics Linden 1.19s, at the median. Both are measured on the same 206 streamed turns.
Speechmatics Linden finishes the transcript 5ms 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. Soniox STT-RT v5 shows first text after 3.92% and Speechmatics Linden after 3.93%, at the median.
Cekura runs these simulated calls on your agent and scores every one.