Tips for the best voice clone quality
Recording environment, speech style, file requirements and fixes for common voice cloning problems.
Written By Umakhan Magomedov
Last updated 29 days ago
Voice clone quality depends on the recording you provide. These tips help you get a natural-sounding result that closely matches your voice.
Recording environment
Record in a quiet room with no background noise, echo or music
Avoid rooms with hard surfaces that create echo (bathrooms, empty rooms)
A small room with soft furnishings (carpet, curtains, sofa) works well
Keep your phone or microphone at a consistent distance from your mouth (15-30 cm)
How to speak
Speak at your natural pace, as you would in normal conversation
Pronounce words clearly without exaggerating
Read the reference text provided in the app — it is designed to capture a wide range of your voice characteristics
Do not whisper or speak unusually slowly: the model learns your natural voice
Aim for at least 30-60 seconds of clean, uninterrupted speech
ℹ️ The reference text in the app is specifically chosen to include a variety of sounds, intonations and sentence structures. Reading it fully gives the model more to work with.
If you upload a file instead of recording
Use a file with a single speaker and no background music or effects
Minimum: 10 seconds. Recommended: 30-60 seconds or more
Maximum file size: 20 MB
Supported formats: MP3, WAV, AAC, OGG
Avoid phone call recordings, heavily compressed audio or recordings with multiple speakers