Appearance
Voice Cloning
Voice Cloning makes a digital copy of your own voice, or of someone who agrees to it on the recording. The person whose voice it is must read a consent statement at the start of the recording.
Clone a voice
Step 1: About the voice
- Open Voice Cloning in the sidebar.
- Fill in:
- Voice name, for example "My narration voice"
- Speaker's full name: the person whose voice this is
- Voice type (optional): Female or Male
- Description (optional)
- Tick the confirmation: "I am speaker, or I have their permission to clone their voice. I won't use it to impersonate, mislead or harm anyone."
- Choose Get the consent statement.
These details lock once you have the statement.
Step 2: Record the voice
- The statement appears under Read this aloud first. It looks like: "My name is Speaker's name. I agree to let Cinara create a digital copy of my voice. My one-time code is four words."
- Under Then keep reading there's a passage to read after the statement. Choose Another passage for different text.
- Choose Start recording. The speaker reads the statement, then keeps reading in their normal voice, without stopping.
- A bar shows progress: "Keep reading: 0:12 to go".
- When it says "Long enough. Stop when you finish the sentence, or keep reading.", you can stop.
- Choose Stop. Listen to the recording. Choose Record again if needed.
- Or choose Upload a recording to use an audio file. It must start with the statement.
- Optionally, tick Reduce background noise in the recording.
- Choose Create voice.
Other controls:
- New statement: get a fresh statement. You'll need to record again.
- Statement valid until time: shows when the statement expires.
- Pick the microphone and choose Test mic before you record.
Step 3: Wait for the check
- You see Step 1 of 2 · Checking the consent statement…, then the voice is created.
- Checking takes a few seconds; creating the voice about half a minute.
- When it's ready: "Name is ready. Consent was verified: N% of the statement heard, one speaker."
- Choose Play preview, Use in Speech or Use in Dialogue. Choose Clone another voice to start again.
If the check fails, you see why, and your credits are refunded. Under the message you see what was heard, for example "Heard 80% of the statement · code not heard · 1 speaker · 0:25 of speech". Choose Try again. See Troubleshooting.
Your voices
The panel on the right lists your cloned voices, with a count out of 10. For each voice:
- Play: hear its preview.
- Use: open Speech.
- Sing: open Singing Voice for this voice.
- ×: delete the voice. It can't be used again, and its recording is deleted.
Cloned voices are also listed in Voice Library under My voices, with Consent verified date.
Where you can use a cloned voice
Your cloned voices are listed first in Speech, Dialogue and Voice Changer. They also appear under Your voices in Studio, Talking Avatars and dub edits, and in Singing Voice.
Limits
- Up to 10 cloned voices per workspace. When full, you see "This workspace has 10 cloned voices, the most allowed. Delete one to make another."
- Recordings: at least 20 seconds of speech, up to 5 minutes. Audio files only.
- A statement expires 30 minutes after you get it, and works for one recording.
- Only one person may be heard in the recording.
- Voices of well-known people can't be cloned.
- Voice name up to 40 characters; speaker's name 2 to 60; description up to 120.
Credits
- 10,000 credits per voice.
- If the consent check or cloning fails, your credits are refunded.
- Using the voice afterwards costs the normal credits of the feature, for example Speech credits per character.
Tips
- Record somewhere quiet, close to the microphone.
- Read the statement word for word, including the code words, then keep going without a long pause.
- Speak naturally, in the voice you want copied.
- Record a little more than 20 seconds. The check counts speech it hears, not the file's length.
- Only tick Reduce background noise in the recording when there is noise.
Related
- Safety and consent
- Singing Voice
- Voice Design: a new voice without cloning anyone.
- Voice Library