Dubbing several characters alone
A three-voice story, a dialogue, a fiction podcast: you act every line, one target voice per character, one node per character. The acting is yours everywhere, only the timbre tells the roles apart.
Change your voice with AI
A line spoken by a male voice, replayed by ElevenLabs Voice Changer in three voices, a woman, a young man, an older man, with the original rhythm and intonation. The full workflow, no account needed.
Interactive demo
Real workflow, real results. Pan and zoom freely.
Use this workflowThe real workflow: the same line enters three Enhance audio nodes, each set to a target voice, and you hear what changes and what stays.
At a glance
- What the workflow produces
- 3 processed audio files
- Canvas structure
- 4 nodes wired by 0 connections
- Models wired in
- ElevenLabs Voice Changer
- Cost of one full run
- about 27 credits, roughly $0.32
A character to dub when you only have one voice at hand, a voice-over you want deeper or younger, a voice message to make anonymous: the voice changer does all of that from an ordinary recording. This workflow runs it through ElevenLabs Voice Changer with three target voices side by side, to hear at once what it changes and what it keeps.
What it keeps is the essential part: the rhythm, the pauses, the hesitations, the intonation. The changer does not rewrite the line, it replays it with another timbre. The demo line has a question, a pause and a punchline, and the three voices reproduce them identically. This is not speech synthesis: it is your acting, in another throat.
Twenty-one ElevenLabs voices to choose from, and no personal voice: the tool disguises, it does not clone. Your file enters through the node's import button, up to ten minutes per run, by the second: the twelve-second line in three voices costs thirty-two cents, nine credits per voice.
How it works
- 1
Record or import the line
A phone voice memo is enough. Speak the way you want it spoken: the changer copies the acting, not the words. A clean take gives a clean result; a noisy background gets cleaned first.
- 2
Pick a target voice
Twenty-one voices, women and men, young and old, in the node's selector. Each node carries its own: three nodes on the same source, three candidates to compare.
- 3
Listen to the rhythm, not the timbre
The timbre will be the chosen voice's, that is a given. What you judge is the faithfulness of the acting: is the pause before the punchline still there, does the question still rise?
- 4
Send to the edit or download
The node's output cables into the montage or into another treatment. For dubbing, one voice per character, each in its own node.
Variations and use cases
The same canvas, tuned for different needs: each variation is two or three lines of prompt away.
Making a testimony anonymous
A testimony you want to broadcast without the person being recognised. The voice changes entirely, the emotion and the rhythm stay: more honest than a slowed or distorted voice, and far more listenable.
A deeper or calmer voice-over
Your voice sounds too young or too high for a narration: replay it with George or Roger. You keep your diction and your silences, you gain the timbre you do not have.
A disguised voice message for a joke
The lightest case, and the most frequent. A birthday message said by a child's or an old man's voice, with your words and your laughs. Say it is a game: the voice is convincing.
An assistant or mascot voice
A brand character that speaks in videos: record the lines yourself, always run them through the same target voice, and the character keeps its voice from one video to the next.
The same text read by a machine
If you have no recording at all, the voice-over workflow is what you need: it generates the reading from the text. The changer always starts from a human voice.
Common mistakes
Expecting a clone
The changer does not reproduce a voice given as an example. Twenty-one voices, all licensed by ElevenLabs, no real person. Wanting "the voice of" someone is outside what the tool does, and outside what it should do.
Changing a voice over a noisy background
The noise is not transformed, it is kept as is under a voice that did change: the contrast makes it more audible than before. Clean first, change next.
Speaking flat and counting on the target voice
The changer does not act for you: a monotone reading gives a monotone target voice. All the acting has to be in the source recording, that is what drives it.
Running a song through it
The changer is made for speech. On singing it sometimes holds the note, often not, and the music behind comes out damaged. For a song, stem separation comes first.
Recommended models
ElevenLabs Voice Changer
DeepFilterNet 3Frequently asked questions
Can I use my own voice as the target?
No, and that is a choice. The changer replays your recording with one of the licensed ElevenLabs voices; it does not learn a new voice and does not reproduce a real person's. For disguising, yes; for imitating someone, no.
Does the voice keep my accent?
It keeps your pronunciation and your intonation, so largely your accent: the changer does not touch the way words are said, only the timbre. A strong accent stays strong, in another voice.
What if the recording is noisy?
The noise passes into the new voice and stands out even more, because the timbre changed and it did not. Clean first with the clean up a recording workflow, then change the voice.
How much does a voice change cost?
Forty-two credits per minute, by the second. The demo line, twelve seconds in three nodes, costs twenty-seven credits, nine per voice. The price shows on the node before you click, on the measured duration of your file.
What is the maximum duration?
Ten minutes per run. A long recording is cut into pieces in the edit, which also lets you re-run only the passage you correct.
Which languages does it work in?
All of them: the changer does not understand words, it transforms sound. A line in Japanese or Arabic comes out in Japanese or Arabic, with the same pronunciation.
Can I do it from my phone?
Yes, and it is even the natural case: a voice memo recorded on the phone, imported into the node from the mobile canvas. The creating on mobile guide explains the gestures.
Are my recordings used to train models?
No. What you upload stays in your workspace and is handed to no training, and no voice is learned from your files: the changer only has its twenty-one voices.
Change your voice with AI
Use this workflowFree account, no card required. The workflow opens pre-filled in your canvas.