ElevenLabs Voice Changer
ElevenLabs' speech-to-speech converter — swap the voice in an existing recording while preserving every whisper, laugh, pause and emotional cue of the original performance.
curl https://api.venice.ai/api/v1/audio/queue \
-H "Authorization: Bearer $VENICE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "elevenlabs-voice-changer",
"prompt": "An uplifting cinematic orchestral build with soaring strings"
}'
# Use the returned queue_id with https://api.venice.ai/api/v1/audio/retrieve.
# Call /audio/complete after downloading if needed.Overview
What is ElevenLabs Voice Changer
ElevenLabs Voice Changer (originally Speech-to-Speech, launched November 2023) converts the voice in an existing recording into a different target voice while preserving the original performance — emotion, timing, pacing, intonation, whispers and laughs. Instead of prompting text, you perform the delivery yourself and the model transfers it to another voice.
Using it anonymously on Venice
On Venice, your source audio and conversion requests are handled under the anonymized privacy tier — prompts are not stored, profiled, or used for training. That matters for voice work: actors anonymizing their voice, studios handling unreleased scripts, and creators re-voicing sensitive recordings get the conversion without building a permanent server-side history. You pay per track in credits rather than committing to a subscription.
Specifications
Datasheet
- Maker
- ElevenLabs
- Modality
- Speech-to-speech voice conversion — audio in, audio out
- Open weights
- No — proprietary
- License
- Proprietary
- Released
- November 22, 2023 (as Speech-to-Speech)
- Max segment length
- 5 minutes per conversion (split longer recordings into chunks)
- Languages
- 29 (multilingual v2 models); a dedicated English-only sts model is also available
- Input
- Recorded or uploaded audio of the source performance
- Noise handling
- Optional background-noise removal flag for noisy source recordings
- Privacy on Venice
- Anonymized — prompts not stored
- Available on Venice since
- Jul 2026
Assessment
Strengths and limitations
- Preserves the full performance: emotion, pacing, intonation, breaths, whispers, laughs and cries carry over to the target voice, which text-to-speech prompting cannot reliably direct.
- Gives precise control over delivery: demonstrate exactly where emphasis falls and how a pause lands by performing it, then apply that delivery to any target voice.
- Works across 29 languages with the multilingual v2 models, enabling consistent character voices across languages and recording sessions.
- Can fix or replace specific words and phrases in an existing recording without re-recording the whole take.
- Optional background-noise removal keeps noisy location recordings usable instead of unintelligible.
- Closed and proprietary: no open weights, so it cannot be self-hosted or fine-tuned.
- Not a real-time tool: it converts finished files, so it is the wrong pick for live Discord, game chat or streaming, where low-latency changers like Voicemod win.
- Maximum segment length is 5 minutes: longer recordings must be split into chunks and stitched.
- It converts to an existing target voice; it does not clone a new voice from a sample — that is ElevenLabs' separate voice-cloning feature.
- Not uncensored or open-source; output passes through ElevenLabs' hosted service.
Use cases
What it is good for
- 01Voice actors extending their range — perform once, deliver in many different voices and tones.
- 02Game and animation character voices kept consistent across sessions and languages.
- 03Dubbing and voice-over replacement while keeping the original actor's timing and emotion.
- 04Fixing mispronunciations or flat delivery in an otherwise finished narration take.
- 05Anonymizing a speaker in a recording while keeping natural, human-sounding delivery.
Prompting
Getting better results
Record your source take deliberately — the model transfers whatever you perform, so exaggerate the emotion, pauses and emphasis you want in the output.
Keep each clip under 5 minutes; split longer recordings into natural sentence or paragraph chunks before converting.
If your source audio was captured in a noisy room, use the background-noise-removal option — unfiltered noise is the most common cause of garbled output.
Choose the target voice first and match it to the role: test a narrator voice against a character voice on the same take before committing credits.
To fix a single mispronounced word, re-perform just that phrase and convert the short clip rather than re-processing the whole recording.
For English-only work, prefer the dedicated English sts model; for other languages, use the multilingual v2 path.
Alternatives
How it compares
| Model | Best for | Voices | Open weights | Price (Venice) |
|---|---|---|---|---|
| ElevenLabs Voice Changer | Re-voicing existing recordings | 10,000+ voice library | No | from $0.35 / track |
| ElevenLabs TTS v4 | Script-to-speech narration | Preset voice library | No | $0.09 / 1K chars |
| ACE-Step 1.5 | Music generation | — | Yes | from $0.03 / track |
| ElevenLabs Sound Effects | Sound design from text | — | No | $0 / sec |
Pick ElevenLabs Voice Changer when you already have the recording and the performance is the point — dubbing, character work, delivery fixes. When you're starting from a script instead, ElevenLabs TTS v4 is the better entry point.
API
Call it from your code
Venice exposes this model through the REST API. Queue a generation with the model id.
curl https://api.venice.ai/api/v1/audio/queue \
-H "Authorization: Bearer $VENICE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "elevenlabs-voice-changer",
"prompt": "An uplifting cinematic orchestral build with soaring strings"
}'
# Use the returned queue_id with https://api.venice.ai/api/v1/audio/retrieve.
# Call /audio/complete after downloading if needed.Pricing
What it costs on Venice
Pay per track on Venice — price scales with length, from $0.35.
New Venice accounts include a free daily allowance and 500 welcome credits — no credit card required.
FAQ
Frequently asked questions
ElevenLabs Voice Changer, originally called Speech-to-Speech, converts the voice in an existing recording into a different target voice while preserving the original tone and delivery — emotion, timing, pacing, whispers and laughs all carry over. You perform the delivery, and the model swaps who is speaking.
Venice bills per track, with the price scaling by length: a 60-second clip is $0.35, 120 seconds is $0.69, 180 seconds is $1.04, 240 seconds is $1.38, and 300 seconds is $1.73. There is no subscription required — you pay in credits per conversion.
Neither. It is a proprietary ElevenLabs model with closed weights, so it cannot be self-hosted or fine-tuned, and on Venice it is billed per track rather than offered free. If you need open weights for audio work, ACE-Step 1.5 is the open alternative on Venice.
They solve different problems. Voice Changer starts from an audio recording and transfers your performed delivery — emotion, pauses, emphasis — onto another voice. TTS v4 starts from written text and generates speech directly. Use Voice Changer when the performance already exists in a take; use TTS v4 when you're starting from a script.
The maximum segment length is 5 minutes per conversion. Longer recordings should be split into chunks and converted separately — natural sentence or paragraph boundaries make the seams easiest to stitch.
The multilingual v2 models support 29 languages, including English (multiple variants), Japanese, Chinese, German, Hindi, French, Spanish, Portuguese, Arabic, Korean, Polish, Ukrainian and Russian. A dedicated English-only sts model is also available when you only need English.
Yes, with a caveat. There is a background-noise-removal option that should be enabled for noisy source audio — testing on café-noise recordings has shown output can become largely unintelligible without the filter and clean with it. For best results, start from the cleanest recording you can.
No. Voice Changer converts a recording into an existing target voice; creating a new voice from a voice sample is ElevenLabs' separate voice-cloning feature. If your goal is a brand-new custom voice, look at the cloning tools rather than Voice Changer.
On Venice, the model runs under the anonymized privacy tier: your prompts and conversion requests are not stored, profiled, or used for training. You can convert audio without building a persistent personal history tied to your account.
Use ElevenLabs Voice Changer anonymously
Venice does not store your prompts. Chat history stays in your browser.