conecto. Get started
Speech

Transcribe audio

speech_transcribe · Reads only · Runs in the background · Markdown

Turns the speech in an audio or video file from the store into text, in about 100 languages, with speakers and timestamps if asked.

$0.01 per started minute of audio

From your agent’s balance. A call that fails costs nothing.

An example

Transcribe the recording of Saturday's bike fitting with Rachel Kim, and note who said what.

Your agent callsspeech_transcribe$0.01 per started minute of audio

The example business is Halverson Cycle & Repair, a bike shop with a repair workshop in Portland, Oregon. The reply is the one the tool really gives.

With these arguments
{
  "file_id": "7c9e1b3d-5f7a-4c9e-8b1d-3f5a7c9e1b4d",
  "speakers": true,
  "language": "de"
}
What comes back
{
  "text": "Hi, thanks for coming in. My knees hurt after about twenty miles, so I think the saddle is too low. Let us raise it by a centimeter and move it back a little, then ride a few minutes. Okay, that already feels better.",
  "language": "eng",
  "duration_seconds": 15.2,
  "minutes_charged": 1,
  "segments": [
    {
      "speaker": "speaker_0",
      "start": 0.4,
      "end": 3.1,
      "text": "Hi, thanks for coming in."
    },
    {
      "speaker": "speaker_1",
      "start": 3.6,
      "end": 7.9,
      "text": "My knees hurt after about twenty miles, so I think the saddle is too low."
    }
  ]
}

Use it when

A recording should become text: a voicemail, a meeting, an interview, a video.

  • For the business's own phone calls, transcribed already, use phone_get_call.

What it does

Turn the speech in an audio or video file from the file store (MP3, WAV, M4A, OGG, FLAC, WebM, MP4, MOV, up to 25 MB) into text, in any of about 100 languages, detected by itself. With speakers, it says who spoke when; with timestamps, it gives timed segments. Runs in the background: poll the call_id. Charged per started minute of audio, never more than max_minutes.

What comes back

When it finishes: the text, the language, duration_seconds, minutes_charged, and with speakers or timestamps the segments (speaker, start, end and text). Every reply also carries cost_usd and balance_usd.

Inputs

NameTypeWhat it is
file_idRequiredstringExactly 36 characters.
languagestringISO 639 code like "en" or "deu" when you know it. Matches ^[a-z]{2,3}$.
speakersbooleanDefault: false.
timestampsbooleanDefault: false.
max_minutesintegerThe most minutes to pay for; audio past it is transcribed but not charged. 1 to 180. Default: 60.

The call

Over MCP, call speech_transcribe on https://api.conecto.chat/mcp/speech. Over REST:

Terminal
curl -X POST https://api.conecto.chat/v1/tools/speech_transcribe \
  -H "Authorization: Bearer $CONECTO_KEY" \
  -H "Content-Type: application/json" \
  -d '{"file_id":"7c9e1b3d-5f7a-4c9e-8b1d-3f5a7c9e1b4d","speakers":true,"language":"de"}'

When it refuses

A refused call costs nothing and says what to do next.

CodeWhenWhat to do
not_found404No file in the store has that file_id.Find it with files_list.
not_audio400The file is not audio or video.Upload the recording with files_upload first.
unreadable422The speech service could not read the file.Send a common audio or video format.
unavailable503Speech is not switched on yet, or the speech service failed or did not answer.Try again later; nothing was charged.

More Speech tools

All 3 tools

Give this to your agent.

Set up Conecto from https://conecto.chat/SKILL.md