Audio intelligence

Turn every recording into data you can use.

Transcripts, then summaries, entities, sentiment, redaction and translation, asked for in the same call or added later.

Start with 1 free hourRead the docs
1 h
Free to start
96
Languages
9
Features

Transcribing· ENT consultation · Kiswahili and English

01:09
  • 00:00 Nice to see you today.
  • 00:46 Haya yanawasha tu hayaumi, hayaumi hakuna kitu.
  • 01:09 Mara nyingi ni dalili ya kwamba pia una allergy.
Example outputDraft note · ear itching, no pain

Features

Do more with every recording.

Each card shows what the feature does to a line of text.

  • Included

    Transcription

    Timestamped text in 96 languages, including speakers who switch languages mid-sentence.

  • Included

    Custom vocabulary

    Pass up to 25 names, drugs or terms with a job so they come out right.

  • Add-on

    Translation

    Any language into any other, line by line, timestamps kept.

  • Add-on

    Summary

    Three to five bullet points from a long recording.

  • Add-on

    Named entities

    People, organisations, places, dates, medicines and numbers as structured data.

  • Add-on

    Sentiment

    Positive, neutral or negative for each timed line, plus the overall tone.

  • Add-on

    PII redaction

    Emails, phone numbers and ID numbers masked by fixed rules, before the text leaves you.

  • Preview

    Live transcription

    Speak or stream a file and lines arrive a few seconds after each pause. In the console and over a WebSocket.

  • Beta

    Speaker labels

    Who said what: each line labelled Speaker A, B and so on. Turn it on per job.

Connectors

Every feature, wherever your audio lives.

Recordings arrive from Zoom, Teams, Meet, Drive, Dropbox, OneDrive, S3 and Twilio; results go to Slack, email or your webhooks.

Meetings

  • ZoomCloud recordings, as they finish
  • Google MeetThe Meet Recordings folder
  • Microsoft TeamsRecordings from OneDrive
  • Meeting botJoins any call and records it

Files

  • Google DriveWatches a folder
  • DropboxWatches a folder
  • OneDriveWatches a folder
  • Amazon S3Reads a bucket, writes results back

Calls and results

  • TwilioCall recordings, as they finish
  • SlackPosts each summary to a channel
  • EmailSends each transcript to a list
  • WebhooksSigned events to your own app

AI assistants

An MCP server: search, transcribe, translate and summarise from your assistant, with an API key.

  • Claude
  • Cursor
  • Any MCP client

AI assistants

Ask your recordings, from any AI assistant.

ADB Voices is an MCP server. Connect Claude, Cursor or any MCP client with an API key, and it can find what was said, start transcriptions, translate and summarise, inside your workspace and your key's permissions.

  • search_transcriptsFind recordings by words spoken
  • get_transcriptRead a transcript, with speakers
  • transcribe_urlStart a job from a link
  • translate_transcriptInto any language
  • analyze_transcriptSummary, entities, sentiment, redaction
  • get_balanceMinutes left in the workspace
mcp.jsonAny MCP client
{
  "mcpServers": {
    "adb-voices": {
      "url": "https://api.voices.adbtech.ai/mcp",
      "headers": {
        "Authorization": "Bearer <your API key>"
      }
    }
  }
}
“What did the minister say about fuel prices last week?”The assistant calls search_transcripts, then get_transcript, and answers with the timestamps.

How it works

Ask for it with the transcript.

  1. 1

    Send the recording

    One call with the file and the features you want. Add dry_run to see the price first.

  2. 2

    Or add them later

    One call on a finished transcript adds translations or features.

  3. 3

    Get structured results

    JSON your product can use, with the cost of each item on the job.

POST /v1/transcriptions
file=@visit.m4a  language=sw  translate_to=en
features=summary,entities,sentiment,redact

GET /v1/transcriptions/tr_…
→ { "status": "completed",
    "features": {
      "summary":  { "result": { "bullets": ["Patient reports ear itching for two weeks", …] } },
      "entities": { "result": { "entities": [{ "text": "two weeks", "type": "date" }, …] } }, … },
    "cost": { "kind": "final", "total": "0.018500", "lines": [ … ] } }

Pricing

Core included. Add-ons when you need them.

Transcription and custom vocabulary are included. Add-ons are billed per hour of audio, only when you turn them on.

See pricing

FAQ

Straight answers

Do add-ons cost extra?

Transcription and custom vocabulary are included. Translation, summary, entities, sentiment and PII redaction are billed per hour of audio, only when you use them.

Can I run it more than once?

Each add-on runs once per recording. Asking again returns the stored result at no charge.

Does redaction use AI?

No. It is rule-based, so it is predictable and nothing is sent to a model.

Does it work in a sovereign deployment?

Yes. Every feature runs on the same model inside your own servers.