ElevenLabs Voice Changer
What is ElevenLabs Voice Changer?
ElevenLabs Voice Changer is the speech-to-speech endpoint. Upload audio, pick a target voice, and the same speech comes back delivered in the new voice with the original timing, inflection, and emotional delivery preserved. The required fields are minimal: audio_url for the source and voice_id for the target. Optional fields control the underlying model (model_id defaults to eleven_multilingual_sts_v2), the output container (output_format), input-format hints (file_format), background-noise removal, an integer seed between 0 and 4,294,967,295 for reproducibility, and a voice_settings object with stability, similarity boost, style, speed, and a speaker boost toggle. The point of the endpoint is the trade it lets you make: keep the performance, change the voice.
Key features of ElevenLabs Voice Changer
Five features cover what you'll touch on a typical job.
Performance-preserving voice swap
Original timing, inflection, breath, and emotional delivery survive the conversion. Only the voice itself changes. That is what makes the endpoint useful for re-voicing a take rather than re-recording it.
Background noise removal in the same call
Set `remove_background_noise` to `true` and the model strips noise from the input before conversion. One call replaces a denoise pass plus a voice change pass, and the timing stays locked through both.
Configurable speech-to-speech model
`model_id` defaults to `eleven_multilingual_sts_v2`, which fits most localization and brand-voice work. Override it when a project needs a different ElevenLabs speech-to-speech model.
Voice settings for tonal control
Stability, similarity boost, style, speed, and a speaker boost toggle let you shape the target voice without re-cloning. Run the same source through different settings to A/B brand-voice options.
Seeded, reproducible output
Pass an integer seed (0 to 4,294,967,295) and the same audio, target voice, model, and settings return the same converted audio. Useful for regression tests and locked-down production runs.
Best for
Content localization
Re-voice narration in a consistent brand voice across regions while keeping the original performance and timing.
Podcast guest standardization
Convert guest audio to a polished house voice when the recording quality is uneven, or unify multiple guest voices into one for accessibility builds.
Voiceover replacement for finals
Swap placeholder reads with final talent voice on the same takes. Inflection survives, so the cut does not need to be re-edited around new timing.
Accessibility builds
Convert speech to a clearer or more intelligible target voice for hearing-assistance products, with original cadence preserved.
Audio dramas and games
Transform a single performance into multiple character voices. Useful for solo creators producing radio plays or interactive fiction.
Brand-voice demo reels
Produce multiple voice styles from a single recording, then send the variants to the client for selection without booking new sessions.
Use cases
Localize a hero brand spot by recording the performance once in English, then running it through Voice Changer with target voice IDs for each market. Standardize podcast guest audio by routing every guest interview through a consistent house voice while keeping the original phrasing and emotion intact. Replace a temp voiceover on a finished cut with the final talent voice without re-editing the timeline, since the converted audio matches the original timing frame-for-frame. Build a brand-voice presentation by sending one source recording through five different target voices and letting the client pick a winner.
API examples
Call ElevenLabs Voice Changer from any language by POSTing to /v1/tasks. Full parameter docs live at docs.unifically.com/models/audio/elevenlabs/voice-changer.
curl -X POST https://api.unifically.com/v1/tasks \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "elevenlabs/voice-changer",
"input": {
"audio_url": "https://example.com/source-audio.mp3",
"voice_id": "TX3LPaxmHKxFdv7VOQHJ"
}
}'
Successful submission returns a task_id. Poll GET /v1/tasks/<task_id> or set a callback_url on the request to receive the finished audio URL.
FAQs
People also ask
ElevenLabs Voice Changer is a speech-to-speech endpoint. Upload audio via audio_url, pick a target ElevenLabs voice via voice_id, and the model returns the same speech delivered in the new voice with the original timing and inflection preserved.
Two required fields. audio_url for the source audio. voice_id for the target ElevenLabs voice. Optional fields include model_id, output_format, file_format, remove_background_noise, seed, and voice_settings.
eleven_multilingual_sts_v2. Override it via the model_id parameter when you need a different ElevenLabs speech-to-speech model. The default fits most localization and brand-voice work.
Yes. Set remove_background_noise to true and the model strips background noise from the input before voice conversion. Useful when the source recording was made in a noisy room or on the phone.
TTS reads new text in a chosen voice. Voice Changer keeps the original speech, timing, and inflection but swaps the voice on top. Use Voice Changer when the performance matters and only the voice should change.
Yes. Pass an integer seed between 0 and 4,294,967,295. Re-running the call with the same audio, target voice, model, and settings returns the same converted audio.
Five fields. stability and similarity_boost shape how closely the output matches the target voice. style exaggerates the speaker's character. speed is a speech speed multiplier. use_speaker_boost adds extra clarity at a small latency cost.