Use case · Meeting notes

Speech-to-text for meeting notes

Meeting notes start with a transcript you can trust: names spelled right, numbers written as numbers, and timings you can link back to. Wire-1, Saidwire's English speech-to-text model, handles both sides of the job. Stream a meeting live for notes as it happens, or send the recording when it ends. On Coval's STT benchmark, its word error rate was 3.8% over the 30-day window to Nov 9, 2026, with a mean time to final segment of 37 ms on far-field microphones and 26 ms on accented speech. Source: Coval model page, 30-day window to Nov 9, 2026, 2,857 samples, US-hosted endpoint. Add your own network round trip. Method and how to reproduce.

After the meeting: send the recording

curl -s -X POST "https://api.saidwire.com/v1/listen?punctuate=true&numerals=true&keyterm=Saidwire&keyterm=Q4%20roadmap" \
  -H "Authorization: Token $SAIDWIRE_API_KEY" \
  -H "Content-Type: audio/mpeg" \
  --data-binary @standup.mp3 \
  | jq -r '.results.channels[0].alternatives[0].transcript'

Or with the OpenAI SDK, if your notes tool already uses it:

# pip install openai
import os
from openai import OpenAI

client = OpenAI(base_url="https://api.saidwire.com/v1", api_key=os.environ["SAIDWIRE_API_KEY"])
with open("standup.mp3", "rb") as f:
    r = client.audio.transcriptions.create(model="saidwire-en", file=f)
print(r.text)

Files can be WAV, MP3, FLAC, Ogg/Opus or WebM, up to 100 MB and 60 minutes per request. Each word comes back with start and end times, so each line of your notes can link to the moment it was said.

During the meeting: stream it

Stream the meeting audio to wss://api.saidwire.com/v1/listen and append each final result to the running notes. Interim results let you show the line being spoken. Streams run up to 4 hours. See the live captions page for a relay example.

Options for notes

parameter	what it does
`keyterm=`	people, product and project names, spelled the way you wrote them (repeat once per term, up to 100)
`numerals=true`	"three hundred and twelve" becomes "312", "fifty percent" becomes "50%"
`redact=pci,ssn,phone,email`	masks card numbers, Social Security numbers, phone numbers and emails
`punctuate=true`	sentence case and punctuation

From transcript to notes

Saidwire returns the transcript; your own LLM writes the summary, action items and decisions. Pass the transcript with word timings so each action item can cite its timestamp. Saidwire does not keep your audio or transcripts: audio is processed in memory and never stored.

Cost

120 free hours with no card and no deadline, up to 4 streams at once, then $0.25 per audio hour, billed per second. An hour-long meeting costs $0.25.

Questions

Can it label who said what?

Not yet on the public API. Each stream is one channel, so if each person has their own microphone, open one stream per person and label the lines yourself.

How do I get names spelled right?

Pass them with keyterm= (one per term, up to 100). Finals then use your spelling.

How long can a recording be?

Up to 60 minutes and 100 MB per request. Split longer recordings, or stream them live (up to 4 hours per stream).

Do you store meeting audio?

No. Audio is processed in memory and never stored, and transcripts are not logged.

Questions or stuck? Email support@saidwire.com with the request ID from the failed call.

Pipecat, LiveKit, Vapi, OpenAI and the other names on this page belong to their respective owners. They are named only to describe compatibility; no partnership or endorsement is implied. Coval benchmark figures are from Coval's public model page; Coval does not endorse Saidwire.