Call Transcription

Real-time call transcription and AI summarization in dialnote

Every call on dialnote can be automatically transcribed, summarized, and tagged — no manual effort required. Once you turn on transcription for a phone number, all answered calls get a full text transcript with speaker labels, timestamps, and AI-generated summaries.

How It Works#

dialnote uses Twilio Voice Intelligence to transcribe your calls. Here's the flow:

  1. A call is answered on a number with transcription enabled
  2. Each participant is recorded on a separate audio track during the call
  3. When the call ends, the recording is sent for transcription and speaker segments are merged
  4. AI generates a summary with key points, sentiment, and action items
  5. The completed transcript appears in the conversation timeline

Each speaker is identified separately, so you'll always know who said what. For multi-party calls (like conference calls), every participant gets their own labeled track.

Turning on Transcription#

Transcription is configured per phone number, and it's turned on by default for new numbers. To check or change it:

  1. Go to Settings → Phone Numbers and select the number you want to configure
  2. Scroll to the AI & Intelligence section
  3. Toggle AI Transcription on or off

You'll also see two related options that stay disabled until transcription is turned on:

  • AI Summaries — generates a brief summary with key points, sentiment, and action items after each call.
  • AI Contact Extraction — pulls contact details like name, email, and company from the transcript and queues them for review.

A separate AI Call Tags toggle automatically categorizes calls based on the transcript content — it stays available for owners even before transcription is on (though it needs transcripts to actually run). Tags are grouped into four categories: priority, topic, outcome, and action. You define your own tag set under Settings → AI Call Tags.

Reading Transcripts#

After a call ends, the transcript shows up in the conversation timeline inside your Inbox. Each transcript includes:

  • Speaker labels — color-coded by participant so it's easy to follow the conversation
  • Timestamps — every segment is timestamped, and you can click to jump to that point in the recording
  • Confidence indicators — the system tracks transcription accuracy per segment

For calls with three or more participants, a color legend at the top shows which color maps to which speaker. Consecutive segments from the same speaker are grouped together so the transcript reads naturally.

AI Summaries and Action Items#

When AI Summaries are enabled, each transcribed call gets:

  • A short overview — a 2-3 sentence recap of the call
  • Key points — up to five main discussion points
  • Sentiment — whether the overall tone was positive, neutral, or negative (see Sentiment Analysis for more)
  • Action items — specific follow-ups or tasks mentioned during the call

Summaries are generated by Google's Gemini model right after the transcript finishes. They save your team time — instead of listening to a 20-minute recording, you can scan the summary in seconds and decide if the full transcript is worth reviewing.

Transcription Statuses#

Each transcript goes through a few stages:

StatusWhat it means
Not startedCall hasn't been processed yet
In progressTranscription is actively running
CompletedTranscript is ready to view
FailedSomething went wrong — try re-processing

Most transcripts complete within a few minutes after the call ends. Longer calls may take a bit more time to process. If a transcript shows Failed, the recording is kept, so you can re-process it once any issue clears up.

What's Next#

We use cookies for analytics, ads, and to remember your preferences. Privacy Policy