Cloned voice sounds wrong in Percify: fixes that work

Fix a Percify voice clone that sounds unlike you, mispronounces words or won’t record: sample tips, language limits and the errors you might see.

A Percify voice clone copies the voice in the sample you record or upload, so a noisy, very short or music-filled sample gives a weaker match. Record 5 to 60 seconds of your own clear speech in a quiet room, then generate again. Scripts in languages other than English can be mispronounced, because cloned voices in Voices read every script with English pronunciation.

How voice cloning works in Percify

In Voices at app.percify.io/home/voices, press Create or Clone a Voice. Record with Tap to record (a 5 to 60 second sample of your voice) or press Upload to add an audio file. Write the script and press Generate. Percify cleans up background noise in your sample, then reads your script in the cloned voice.

You’re charged only when the audio is ready, so a failed generation costs nothing.

The clone doesn’t sound like me

Why: the clone can only copy what’s in the sample. Room echo, background noise, music, other people talking or a very short take all make the match weaker.

Fix:

1. Record somewhere quiet with soft furnishings, not a bare room, and keep the microphone at a steady distance.
2. Speak naturally in the tone you want in the result. A flat reading gives a flat clone.
3. Use a longer sample. Recordings need at least 5 seconds, and the recorder stops at 60. Use more of that range when you can.
4. Only your voice. No music, no second speaker, no TV in the background.
5. Generate again. Each generation is a new take, so a second run can sound different.

Want your exact voice in a talking video? Record the script yourself in Clone Yourself. That video uses your real recording, not a clone. See Lip-sync problems.

Words are mispronounced

Scripts not in English: cloned voices in Voices read scripts with English pronunciation, so text in other languages can come out wrong or with a strong accent. For those scripts, record the audio yourself in that language and use it in Clone Yourself.

Names, acronyms and numbers: voice models guess how to say these. Write them the way they sound:

• Spell a name phonetically, such as “Nee-am” for Niamh.
• Write numbers as words, such as “twenty twenty-six”.
• Use punctuation for pauses. Commas and full stops change the pacing.

Recording or upload problems

Message	Fix
Could not access microphone	Allow microphone access for app.percify.io in your browser, then try again.
Record at least 5s	Record a longer sample.
Please upload an audio file	The file you picked isn’t audio. Choose an MP3, WAV or M4A file.
File size must be less than 10 MB	Trim or compress the sample.
Failed to upload audio	Check your connection and try again.
Select a voice or record your own first	Record or upload a sample, or pick a library voice.
Write something for the voice to say	Add a script before you press Generate.
Insufficient credits	See Not enough credits.

When to contact support

If a clean sample of at least 5 seconds still gives a voice that sounds nothing like the recording, email [e-mail]. Include the sample, the generated clip and the script.

Only clone your own voice, or a voice you have clear permission to use. Percify’s Terms forbid impersonation and content that violates anyone’s rights.

Related

Clone your voice

Step-by-step voice cloning in Percify.

Lip-sync problems

Fix talking videos made from your audio.

Upload failed

Audio size limits and formats.

Dub a video

Put speech in another language on a video.