.png)
Ho to record
-
Duration: Record 4 to 5 minutes of audio.
Style: Do not read from a script. Speak naturally, just as you would in a normal conversation or presentation. -
Language: Speak in the exact language you want your final avatar to use.
DIGITIZING YOUR PRESENTER VOICE
Submission guidelines for creating your digital voice.
To build a high-grade, realistic digital voice of your company representative, it is key the input is of high quality. Once the voice is cloned, it can be used for all your messaging.
Please read these instructions carefully
Recording & submitting
-
Do not compress: Never send the audio file via WhatsApp, WeChat, or standard text message, as these platforms heavily compress the file and ruin the quality.
How to send: Export the raw audio file directly from your phone (as a .WAV or high-quality audio file) and send it via email or a file-sharing service (like AirDrop, Google Drive, or WeTransfer).
Recording environment
-
Deaden the room: Choose a quiet room with minimal echo. Small spaces with lots of fabrics (like a walk-in closet or a room with heavy curtains and carpets) work best to absorb bounce-back sound.
-
Microphone distance: Keep your phone a consistent distance from your mouth (about two fists away, or 20cm).
-
Speak at an angle: To prevent harsh breathing sounds or "popping" noises from hitting the microphone, speak slightly past the phone rather than directly into it.
Performance & consistency
-
Be consistent: Choose a tone and stick to it. If you want a high-energy avatar, keep your voice animated throughout. If you want a calm voice, keep it subdued. DO NOT MIX STYKES. Mixing styles will result in an unstable clone.
-
Mind the habits: The AI will clone everything—including your pauses, breathing habits, and filler words (like "um" and "ah"). Speak clearly, confidently, but natural. DO NOT READ as this will be noticeable in the final result.