How it works
In abcVoiceover, every voice in a project's cast is cloned from its own short clip. The model copies more than the sound of a voice: it copies the way the clip was performed, its pace, its energy, its mood. Record a slow, gravelly sample and you get a slow, gravelly character; record a quick, bright one and you get a quick, bright character, from the same throat.
What doesn't change much is your basic vocal colour. Every character will share something of your voice, so lean hard on the things you can change.
What to change for each character
- Pitch: speak a little higher or lower than usual. Small shifts sound natural; big ones sound strained.
- Pace: one character rushes, another takes their time. This carries into the clone strongly.
- Energy: bright and loud, or tired and quiet.
- Attitude: warm, suspicious, bored, nervous. Think about what the character wants, and let it colour the reading.
- Texture: breathier, more nasal, rougher, crisper.
- Accent: only one you can do well and respectfully. A caricature of a real accent is grating to hear, and to the people who have it.
An example cast
| Voice | How to record the sample | Style |
|---|---|---|
Narrator | Your natural voice, steady and warm | Calm |
Captain | Lower and slower, with long pauses, as if nothing surprises her | Balanced |
Recruit | Higher and faster, eager, tripping over the words | Expressive |
Stranger | Quiet, close to the mic, every word deliberate | Calm, with a little more emotion intensity |
Recording the samples
- Write a short passage for each character, in their own words, so you're performing them rather than reading neutral text. Two or three sentences each.
- Warm up into the voice before you record, and record each sample in one go.
- Keep the setup identical: same room, same mic, same distance. Then the differences between characters come from your performance, not your equipment.
- Add each as its own voice, named exactly as the character appears in your script, so pasted lines like
Recruit: Yes, sir!go to the right voice. See formatting dialogue.
Set them apart with styles
Each voice keeps its own settings. Give the excitable character Expressive and the measured one Calm, then open Fine-tune on each: more emotion intensity makes a character bigger, and lower guidance strength loosens how tightly a voice follows its sample's pacing. The voice settings guide explains each slider.
Test with a scene
Write a short exchange where every character speaks, generate it, and listen with your eyes closed. If two characters blur together, push one of them further: re-record its sample with a bigger change in pace or energy, rather than adjusting sliders. Saved lines keep their sound if you replace a clip, so you can keep refining as you go.
Limits worth knowing
- Voices very far from your own, such as a small child or a very deep giant, are hard to get from one performer. Another performer, with their consent, may be the better choice.
- Don't build a character by imitating a real, recognizable person. That's impersonation, and the terms of service rule it out.