What was said, simply written.
Made in-house for the calls we handle every day.
Switching a project over
Before
POST https://[current provider]/v1/listen Authorization: Token <current key>
After
POST https://said.simple.biz/v1/listen Authorization: Token <your Simply Said key>
Same request, same words[] response with speaker labels. The rest of your code stays as it is.
Connect a project in three steps
- Step 1
Ask for an invite
Use your @simple.biz Google account. An admin approves it.
- Step 2
Get an API key
While Simply Said is in research, keys come from the admin. Keep it in your project’s own secret store, as SIMPLY_SAID_KEY.
- Step 3
Change the URL
Point your project at said.simple.biz. Keep your old key for 30 days as a fallback.
What it handles
Phone audio
8 kHz call recordings, GSM .wav, mp3 and other formats, converted on the way in.
Speaker labels
Every word is tagged with who said it, so agent and caller are told apart.
Word timings
A start and end time for every word, so a line can jump to its spot in the audio.
Whole calls
Send the full recording in one request; the transcript comes back in the response.
Audio deleted after each job
Recordings stay on our server only while they are transcribed.
Runs on our own server
Built in-house and hosted by us, at said.simple.biz.
Not yet: live, real-time transcription, and callbacks for background jobs.
Quickstart
Send a recording and get the transcript back, with word timings and speakers.
Send a call
curl -X POST \ "https://said.simple.biz/v1/listen?smart_format=true&diarize=true" \ -H "Authorization: Token $SIMPLY_SAID_KEY" \ -H "Content-Type: audio/wav" \ --data-binary @call.wav
Ask for an invite to get a key.
What comes back (trimmed)
"results": {
"channels": [ {
"alternatives": [ {
"transcript": "Thanks for calling, how can I help?",
"words": [ {
"punctuated_word": "Thanks",
"start": 0.42, "end": 0.71,
"speaker": 0
} ]
} ]
} ]
}