Skip to main content

Voice Fingerprints

Written by Alyssa Elso

What are Voice Fingerprints?

Voice Fingerprints help Practice by Numbers identify which team member handled a phone call. This makes call analytics and scoring accurate for each team member.

PbN identifies the team member on a call in two ways:

  1. Name detection (primary method). AI transcribes the call and listens for an introduction such as "Hi, this is Sarah from the office." It then matches that name to your staff records. With clear introductions, this method succeeds more than 90% of the time.

  2. Voice fingerprint (backup method). If it doesn't detect a name, PbN compares the voice on the call to each team member's voice fingerprint. It removes background noise and silence, analyzes voice characteristics, and calculates a similarity score to find the match.

Because the fingerprint is the backup when someone forgets to introduce themselves, every team member who answers phones should have one.

Where to find Voice Fingerprints

Go to Settings > Voice, then select Voice Fingerprints under Voice Management. You can also type "Voice Fingerprints" into the settings search bar.

Here's what you'll see on the page:

  • How It Works opens a quick overview of how voice recognition works.

  • Refresh reloads the list so you can see the latest statuses.

  • Create Voice Fingerprint adds a fingerprint for a team member.

  • Filter by status shows fingerprints by status: All, Ready, Training, Pending, Needs Samples or Failed.

  • Convert All To Modern upgrades all eligible Legacy fingerprints to the Modern profile type. The number in parentheses shows how many can be converted.

  • The fingerprint list shows one row per fingerprint. Each row shows the team member's name, type (Modern or Legacy), status, profile quality percentage, and how many samples have been processed out of the total. Use the microphone icon to add samples and the trash icon to delete the fingerprint.

Practice by Numbers Voice Fingerprints settings page showing the status filters, Convert All To Modern button and list of staff voice fingerprints

Fingerprint statuses

Status

What it means

Ready

The fingerprint is trained and being used to identify this team member.

Training

PbN is processing the samples and building the voice profile.

Pending

The fingerprint has been created and is waiting to be processed.

Needs Samples

There isn't enough usable speech yet. In the list this may show as INSUFFICIENT_SAMPLES. Add more samples.

Failed

The fingerprint couldn't be built. Try adding new, higher-quality samples.

How to create a voice fingerprint

  1. Click Create Voice Fingerprint.

  2. Under Fingerprint Type, choose one of these:

    • Modern ECAPA voice profile is the newer profile type. If the team member already has Legacy samples saved, PbN converts them. Otherwise, it creates an empty profile that you add samples to.

    • Legacy Resemblyzer fingerprint is the original fingerprint type.

  3. Under Employee, select the team member. The Name field fills in automatically.

  4. Optional: add a Description.

  5. Click Create Modern Profile (or Create Legacy Fingerprint).

Create Voice Fingerprint window in Practice by Numbers with Fingerprint Type, Employee, Name and Description fields

The new fingerprint shows up in the list. Next, add voice samples so PbN can train it.

How to add voice samples

  1. Find the team member in the list and click the microphone icon.

  2. The Add Modern Profile Sample window opens. It has two tabs: Record Audio and Upload Files.

Add Modern Profile Sample window in Practice by Numbers showing English and Spanish voice recording text categories

Option 1: Record audio

  1. On the Record Audio tab, choose a set of texts to read: English – Verification Paragraphs, English – Natural Speech, or Español – Verificación de Voz.

  2. The Text to Record box shows the first passage. Click Start Recording and read the passage aloud in your normal speaking voice.

  3. Click Next to move to the next passage. Recording Progress shows how far along you are.

  4. Record at least 2 sample texts. Then finish and save your recordings.

Recording a voice fingerprint sample in Practice by Numbers with the Text to Record passage and Start Recording button

Option 2: Upload audio files

  1. Select the Upload Files tab.

  2. Click Select Files and choose at least 1 audio file of the team member speaking. Each file can be up to 300 seconds (5 minutes) long.

  3. Click Upload.

Upload Files tab for adding voice fingerprint audio samples in Practice by Numbers

Reviewing a fingerprint's details

Click the arrow next to a team member's name to expand the row. You'll see:

  • Profile quality: a percentage that shows how strong the voice profile is.

  • Clean speech: the total seconds of usable speech in the profile.

  • A table of samples with these columns: Sample, Status, Original (length), Usable Speech, Quality, Added (date), Type, and Actions.

Expanded voice fingerprint in Practice by Numbers showing profile quality, usable speech and the voice sample table

What is usable speech? It's the audio PbN keeps for voice matching. The profile builder removes silence, ringback and hold audio, short fragments, very quiet sections, clipping, and noisy sections. A long phone call can produce much less usable speech if only part of it is clear speech from the team member. If a sample's usable speech is below the target, PbN flags it in the table.

Converting Legacy fingerprints to Modern

If you have Legacy fingerprints, click Convert All To Modern to queue all eligible Legacy fingerprints for conversion. The converted profiles appear in the list with the Modern tag. Your Legacy fingerprints stay in the list too.

Deleting a fingerprint

Click the trash icon on the fingerprint's row and confirm. Delete fingerprints for team members who no longer answer phones for your practice.

Best practices

  • Introduce yourself on every call. Name detection is the most reliable way PbN identifies who handled a call. A simple "Thank you for calling, this is Maria" works well.

  • Enroll everyone who answers phones. All phone agents need a voice fingerprint as a backup.

  • Add 3–5 high-quality samples per person. PbN judges sample quality by signal-to-noise ratio, voice clarity, and sample duration.

  • Record in a quiet space. Background noise, hold music, and long silences lower the usable speech in a sample.

  • Speak naturally. Read at your normal pace and tone, the way you would on a call.

Frequently asked questions

  1. My fingerprint says Needs Samples (or INSUFFICIENT_SAMPLES). What should I do?
    The profile doesn't have enough usable speech yet. Click the microphone icon and record or upload more samples. Clear recordings made in a quiet room help most.

  2. Why is a sample's usable speech so much shorter than the recording?
    PbN removes silence, hold audio, noise and very quiet sections before using a sample. If much of the recording was background noise or another speaker, only a small part counts.

  3. Should I use Modern or Legacy?
    Modern ECAPA voice profiles are the newer type. If you still have Legacy fingerprints, use Convert All To Modern to upgrade them.

  4. Can I record samples in Spanish?
    Yes. Choose Español – Verificación de Voz when you record.

  5. A team member left the practice. What should I do?
    Delete their fingerprint with the trash icon so their voice is no longer used for matching.

Did this answer your question?