Deepfake Voice Phishing

Custom voices let you clone a real voice and use it in a simulated voice phishing scenario. This makes training far more realistic by showing employees how convincing modern voice-based social engineering can be.

Attackers are increasingly using AI voice cloning to impersonate executives, co-workers, vendors, and trusted contacts. A familiar voice can lower suspicion, create urgency, and make a phishing attempt feel legitimate. By using custom AI voices in your training, you can better prepare users for the kinds of attacks they may face in the real world.

Upload a Custom Voice

To upload a cloned voice, start in the main Voice Phishing section: 

1. Click AI Voice Library.
2. Click Upload Custom Voice.
3. Choose how you want to provide the source audio:
   • Upload an MP3 or WAV file, or
   • Record audio directly using your microphone
4. Once the audio has been added, upload the voice.

After the upload is complete, the cloned voice will be ready to use in a voice phishing simulation.

Assign the Custom Voice to a Simulation

Once the cloned voice has been created, you can attach it to a scenario.

1. Find the scenario you want to edit
2. Click Update.
3. In the AI Voice section, select Custom.
4. Open the dropdown menu and choose the newly created cloned voice.
5. Click Save Template

The selected voice will now be used in that voice phishing simulation.

Best Practices for More Realistic Voice Cloning

To get the most out of voice cloning, it helps to use source audio that is clear, natural, and at least one minute long. This is not a strict requirement, but longer, cleaner samples usually give the AI model more to work with, which can lead to more realistic speech. Audio with a good range of tone, pacing, and inflection can also improve the result, especially if the speaker sounds natural rather than overly scripted or monotone. If you are recording your own voice, we provide a script you can read to help you create a suitable sample for cloning.

Another way to improve realism is to choose an AI Persona Name and tone in the phishing scenario that matches the cloned voice. For example, if the voice sounds calm and professional, selecting a tone that reflects that will usually produce a more believable result. When the voice, delivery style, and scenario all align, the simulation tends to feel much more convincing.

You can also strengthen the simulation by adding an AI Knowledge Source. This gives the AI extra context about your organization by drawing on public information and any details you provide. If the recipient challenges the caller or asks follow-up questions, the AI can use that context to respond in a way that feels more credible and relevant. This adds another layer of realism and helps the simulation better reflect how a real attacker might use background research to make a voice phishing attempt more persuasive.