Zoom Scribe Pro v2.1

Accuracy vs. speed

Lower and further left is better · Nemotron 3.5 Transcribe doesn't support streaming, so it is not shown

Best quadrantBest picks
Final-text delay P50 (ms)
Overall WER (%)

Zoom Scribe Pro v2.1 is 1st of 24 on the Pipecat Dataset and 14th of 23 on final-text delay.

2.35% Overall WER (5th) · 1.55% Pipecat WER (1st) · 3.87% Ocular WER (13th) · 793ms First text (8th) · 178ms Final text (14th)Likely range, 95% confidence: Pipecat WER 1.35 to 1.75%, Ocular WER 3.14 to 4.67%Published price $0.40 per hour of streamed audio. Scribe API Pro (Live) rate, paid from prepaid credits. Zoom lists no separate rate for v2.1. Source, checked Oct 8, 2026.

Last updated Methodology by Dileep Chagam

Zoom Scribe Pro v2.1 against every other model

Lower is better on every measure. Hover a column for that model's full numbers.

  1. 1.5%
    Scribe Pro v2.1
  2. 1.7%
    STT 3.7 Realtime Agent
  3. 1.7%
    Rumble-2
  4. 1.8%
    Universal 3.6 Pro
  5. 1.8%
    Scribe v3 Live
  6. 1.9%
    Nemotron 3.5 Transcribe
  7. 1.9%
    Transcribe Stream 1
  8. 1.9%
    Universal 3.5 Pro
  9. 2.0%
    Chirp 3
  10. 2.1%
    Resonant-1
  11. 2.2%
    Realtime Whisper
  12. 2.4%
    Ink 2
  13. 2.5%
    Linden
  14. 2.5%
    STT-RT v5
  15. 2.6%
    Chirp 2
  16. 2.7%
    STT-1
  17. 3.2%
    Finch-Global
  18. 3.4%
    4o Mini Transcribe
  19. 3.5%
    Nova-3
  20. 3.7%
    Pulse
  21. 3.9%
    4o Transcribe
  22. 3.9%
    Flux English
  23. 4.0%
    Gradium
  24. 5.0%
    Flux Multilingual
Glossary: WER, TTFT, TTFS, P50
WER · Word error rate
Share of words the model got wrong: replaced, missing or added. Lower is better.
Pipecat Dataset
1,000 short public clips from Pipecat's STT benchmark dataset. Each clip is one turn.
Ocular Dataset
300 turns from 4 real phone conversations, licensed from Ocular. 206 of them are timed.
Likely range · 95% range
Where the model's true score probably falls, given how much audio was tested. Overlapping ranges are too close to call.
TTFT · Time to first text
How long after the audio starts streaming until the first words appear.
TTFS · Final-text delay
How long after the caller stops until the full transcript arrives. A voice agent waits for this before it replies.
P50 / P90 / P95 · Typical, slow and slowest turns
Half of turns are faster than P50. Only 1 in 10 is slower than P90, and 1 in 20 slower than P95.

Where Zoom Scribe Pro v2.1 struggled

On real calls, dropped words are its biggest source of error. On clean clips, wrong words lead. Our guide explains how word error rate is measured.

Wrong words

What the speaker said, with Zoom Scribe Pro v2.1's mistakes marked: I want to start composting written as composing my kitchen scrap to reduce waste, but I live in an apartment with no yard, so I'm looking for a good indoor composting written as composing solution

    Everything Zoom Scribe Pro v2.1 wrote

    I want to start composing my kitchen scrap to reduce waste, but I live in an apartment with no yard, so I'm looking for a good indoor composing solution.

    Pipecat Dataset clip pipecat-e306c9ad-0cbd-d23e-9477-2b938925730c

    Wrong word

    What the speaker said, with Zoom Scribe Pro v2.1's mistakes marked: I am trying to learn Spanish and written as I would like you to translate the phrase where is the nearest train station and then say it slowly so I can practice my pronunciation

      Everything Zoom Scribe Pro v2.1 wrote

      I am trying to learn Spanish. I would like you to translate the phrase, where is the nearest train station? And then say it slowly so I can practice my pronunciation.

      Pipecat Dataset clip pipecat-d6978d46-c6ea-b7d9-0423-e0008303c7d8

      Marked: words Zoom Scribe Pro v2.1 left out or got wrong. Audio from the Pipecat Dataset, under CC BY 4.0, re-encoded as MP3.

      TTFT and TTFS at P50, P90 and P95

      Timed on the same 206 streamed Ocular turns as every other model. P90 and P95 show what a caller meets on a slow turn.

      TTFS, final text

      How long after the caller stops until the full transcript arrives. A voice agent waits for this before it replies.

      10ms50ms200ms1s
      Typical, P50
      178ms
      14th fastest of 23
      Slow, P90
      735ms
      19th fastest of 23
      Slowest, P95
      955ms
      19th fastest of 23

      TTFT, first text

      How long after the audio starts streaming until the first words appear.

      200ms1s5s20s
      Typical, P50
      793ms
      8th fastest of 21
      Slow, P90
      1.19s
      9th fastest of 21
      Slowest, P95
      1.25s
      7th fastest of 21

      Every other model's typical turn, on the same axis.

      1,000 Pipecat turns · 300 Ocular turns · 206 turns timed · 0 failed requests

      Zoom Scribe Pro v2.1 FAQ

      • Zoom Scribe Pro v2.1 places highest on the Pipecat Dataset, 1st of 24 at 1.55%. Its weakest placing is final-text delay, 14th of 23 at 178ms.

      • Zoom Scribe Pro v2.1 transcribes the 1,000 public Pipecat clips at 1.55% word error rate and the 300 licensed Ocular turns at 3.87%. Word error rate counts incorrect, missing, and extra words against the reference transcript, so lower is better.

      • First text arrives after 793ms at the median, and the transcript is final 178ms after the speaker stops. Both are measured on the same 206 streamed turns as every other model, and the page shows P90 and P95 as well, which is what a caller meets on a bad turn.

      • Every model runs the same two datasets under the same conditions: 1,000 public clips from Pipecat's STT benchmark dataset for accuracy at scale, and 300 turns of licensed Ocular conversations for real speech. Latency comes from 206 streamed turns, using each provider's own supported finalization contract. Zoom Scribe Pro v2.1 returned usable text for 1,000 of the public clips.

      Building on Zoom? Test your agent the same way.

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