ElevenLabs Voice Remixing is a feature that applies text-prompted transformations to an existing voice, modifying its core attributes while attempting to maintain the recognisable characteristics that define it. Where Voice Design creates entirely new synthetic voices from scratch, Voice Remixing starts with a voice you already have and modifies it — adjusting gender, accent, speaking style, pacing, and audio quality through natural language instructions.
The feature launched in alpha in 2026 and is now available on the core ElevenLabs platform and via API. It is positioned as a complement to Voice Design and Voice Cloning rather than a replacement — the three tools serve different workflow needs in ElevenLabs’ voice creation ecosystem.
Which Voices Can Be Remixed
Voice Remixing only works with voices that belong to your account — you cannot remix voices from the ElevenLabs Voice Library. The compatible voice types are:
- Instant Voice Clones (IVCs) — voices you created from short audio samples.
- Professional Voice Clones (PVCs) — high-fidelity voice replicas from 30-minute recording sessions.
- Voice Design voices — synthetic voices you created from text prompts.
- Default voices — ElevenLabs’ pre-loaded voices (note: these expire December 31, 2026).
Voice Library voices (voices shared by other ElevenLabs users and added to your collection) cannot be remixed. If you want to modify a Library voice, you would need to first clone it or create a Voice Design voice with similar characteristics, then apply remixing to your own version.
Prompt Strength Levels
The prompt strength parameter is the most important control in Voice Remixing. It determines how aggressively the model transforms the source voice in the direction specified by your prompt:
| Strength Level | Effect | Best For | Risk |
| Low | Subtle modifications — most original characteristics preserved | Minor accent adjustments, slight pacing changes, audio quality improvement | Safe — voice remains recognisably similar to original |
| Medium | Balanced transformation — key attributes modified while preserving voice identity | Accent changes, speaking style shifts, moderate gender adjustment | Moderate — voice character changes but core identity maintained |
| High | Strong adherence to remix prompt — may significantly change tonality | Major accent, gender, or style transformation | Higher — voice may differ substantially from original |
| Max | Full transformation of the voice | Extreme transformations, experimental exploration | High — voice may not maintain recognisable continuity with original |
ElevenLabs’ recommendation: start at Medium for most use cases. If the transformation is too subtle, increase to High. Reserve Max for experimental exploration or when a full voice transformation — rather than a modification — is the goal. Always listen to the remix preview before saving, as the interaction between voice character and prompt strength produces results that are not always predictable from the settings alone.
How to Use Voice Remixing: Step by Step
In the Dashboard
Navigate to Voices in the ElevenLabs dashboard → find the voice you want to remix in My Voices → click the Remix option (indicated by a remix icon) → write your remix prompt in the description box → select prompt strength → click Generate. Three remix previews are generated. Listen to each and select the version that best matches your intent. Save the chosen version as a new voice in your library. The original voice is preserved — remixing creates new versions rather than overwriting the source.
Via the API
The Voice Remixing API endpoint is POST /v1/text-to-voice/remix. Required parameters: voice_id (the ID of the voice you want to remix), voice_description (your remix prompt, 20–1,000 characters), text (preview content, 100–1,000 characters). Optional parameters: prompt_strength (float 0–1, where 0 is Low and 1 is Max), guidance_scale (float — lower values allow more creative interpretation), remixing_session_id (string — link this remix to an existing session for iterative tracking), remixing_session_iteration_id (string — attach to a specific iteration within a session). The response returns three previews with generated_voice_ids and base64 MP3 audio. Save the chosen voice using the generated_voice_id with the /v1/text-to-voice endpoint.
Iterative Remixing: The Most Effective Workflow
The most reliable Voice Remixing workflow is iterative rather than attempting to describe all desired changes in a single prompt. Complex transformations — changing accent, speaking style, AND pacing simultaneously — are less predictable than single-attribute transformations applied sequentially.
Recommended iterative approach
- Step 1: Identify the most important change needed (usually accent or gender — the characteristics that most affect voice recognition).
- Step 2: Write a focused prompt for that single change at Medium strength. Generate and review. If satisfied, save this as ‘Voice v2’.
- Step 3: Remix ‘Voice v2’ with a prompt focused on the second most important change (e.g., speaking style or pacing). Generate and review. Save as ‘Voice v3’.
- Step 4: Repeat for any remaining attributes. Each remix builds on the previous result, accumulating changes progressively rather than attempting them all at once.
This approach produces more controlled, predictable results than single-step complex remixes, and the session ID tracking in the API allows you to maintain a clear lineage of transformations applied to the original source voice.
Effective Remix Prompts
| Use Case | Example Prompt | Recommended Strength |
| Accent change (British to American) | Change the accent to General American — Midwest US, no regional markers, neutral and broadcast-quality | Medium to High |
| Gender shift | Transform to a male voice with similar warmth and pacing — baritone register, mid-30s | High |
| Speaking style — formal to casual | Make the delivery more conversational and relaxed — less formal, slightly faster pace, natural contractions | Low to Medium |
| Audio quality improvement | Improve audio clarity and reduce any recording artefacts — studio quality, clean and bright | Low |
| Age adjustment | Age the voice by approximately 15 years — deeper register, slightly slower and more measured delivery | Medium |
| Emotional character | Add more warmth and approachability to the delivery — friendly, engaging, with a naturally upward inflection | Low to Medium |
| Accent strengthening | Intensify the Scottish accent — more pronounced regional markers, Glasgow area | Medium to High |
Voice Remixing vs Voice Design: When to Use Each
| Scenario | Use Voice Remixing | Use Voice Design |
| You have an existing voice that is 80% right but needs accent adjustment | ✅ — modify without starting over | ❌ — would require rebuilding from scratch |
| You need a completely new voice with no existing starting point | ❌ — requires existing voice to work from | ✅ — generates from text prompt only |
| You want to create gender variations of your Professional Voice Clone | ✅ — modify PVC gender with High strength | ❌ — would not replicate PVC characteristics |
| You need 10 distinct character voices for a game | ❌ — would need 10 source voices to remix | ✅ — generate each character voice from text prompt |
| You want to improve audio quality of an IVC without re-recording | ✅ — Low strength audio quality prompt | ❌ — no existing audio to improve |
| You need a completely different accent but similar core identity | ✅ — High strength accent remix | Either — design may produce a cleaner result |
Related: Complete guide to ElevenLabs Voice Design v3 — creating new voices from text prompts
Default Voice Sunset and Voice Remixing
ElevenLabs’ Default voices expire December 31, 2026. Voice Remixing is one of the migration paths ElevenLabs explicitly recommends for creators who want to adapt existing Default voices rather than completely replacing them. If your production workflow uses Rachel (a warm, professional American female voice), you can remix any Default voice — or a Voice Design equivalent — toward Rachel’s characteristics, then save that remixed voice as a permanent replacement unaffected by the deprecation deadline.
The advantage of using Voice Remixing for Default voice migration (versus simply selecting a Library replacement voice) is character continuity. Your audience or application users have a context for a specific voice character. A remixed voice that preserves the core identity of the replaced Default voice creates less noticeable disruption than switching to a completely different voice from the Library.
Related: Complete guide to ElevenLabs Default Voices expiring December 2026 — migration options
Three Insights Most Voice Remixing Coverage Misses
1. Remixing Works on Voice Design Outputs — Creating Variation Families
Most documentation and guides focus on remixing Voice Clones. Voice Remixing also works on voices created through Voice Design — and this creates a powerful workflow for building voice families. Design a base voice, then create multiple remixed variations: same base character but different accents for multilingual content, same base character but different ages for character series, same base character but different energy levels for different content types. Each variation is a distinct voice with a recognisable family resemblance to the original, produced without any additional recording.
2. The Default Scripts Feature Removes the Prompt Writing Burden
ElevenLabs provides Default Scripts for Voice Remixing — pre-configured scripts optimised for voice remixing workflows. These scripts are specifically designed to expose the characteristics that remix transformations affect most, making it easier to evaluate whether a remix has produced the intended result. Using a Default Script rather than arbitrary preview text gives more consistent and comparable evaluation results across different remix iterations.
3. High and Max Strength Can Produce Genuinely New Characters
ElevenLabs positions Max strength as ‘a full transformation of the voice’ and warns it may change the voice entirely. This framing treats Max strength as a risk. An alternative framing: Max strength with a targeted prompt is actually a fast path to creating new voice characters from an existing voice as a starting point — without the iterative prompt refinement required in Voice Design from scratch. If the remixed voice at Max strength is interesting and usable, even if it sounds nothing like the source, it is a creative output worth saving. Treat Max strength remixing as an exploration tool rather than a precision modification tool.
Voice Remixing in 2027
Voice Remixing is a relatively new feature in alpha as of 2026, and its development trajectory suggests several improvements. The attribute modification capabilities — currently focused on gender, accent, style, pacing, and audio quality — will likely expand to include finer emotional character adjustments, speaking speed controls, and more precise accent targeting within regions (not just ‘British’ but ‘Edinburgh Scottish’ vs ‘Glaswegian’ vs ‘RP’). The iterative session tracking will likely become more visible in the dashboard interface, allowing non-developer users to visualise and navigate the transformation history of a voice without accessing the API. And Voice Library voice remixing — currently limited to owned voices — may open to community voices with appropriate licensing, expanding the pool of starting voices significantly.
Key Takeaways
- Voice Remixing transforms voices you own (IVC, PVC, Voice Design, Default) through text prompts — modifying gender, accent, style, pacing, and audio quality without re-recording.
- Start at Medium prompt strength for most use cases. Use Low for subtle adjustments, High for major transformations, Max only for experimental exploration.
- Iterative remixing — applying one change at a time — produces more controlled results than attempting complex multi-attribute transformations in a single prompt.
- Voice Remixing does not work on Voice Library voices — only voices belonging to your account.
- Use Voice Remixing for Default voice migration: adapt existing Default voice characters into owned, permanent voices before the December 31, 2026 expiry deadline.
Conclusion
ElevenLabs Voice Remixing is the most underutilised tool in the ElevenLabs voice creation ecosystem in 2026. For creators who have built voice identities through cloning or Voice Design and want to extend them — adding accent variations, gender alternatives, or style adaptations — Voice Remixing eliminates the need to start from scratch. The iterative approach produces reliable results even for complex transformations. And with the Default voice expiry deadline approaching, Voice Remixing provides a practical migration path for creators who want to preserve voice character continuity rather than switching to entirely new voices from the Library. Start with a voice you own, write a focused single-attribute prompt at Medium strength, and evaluate — the entire workflow takes under five minutes per variation.
Frequently Asked Questions
What is ElevenLabs Voice Remixing?
Voice Remixing is a feature that transforms an existing voice you own by modifying its gender, accent, speaking style, pacing, or audio quality through natural language prompts. It works with Instant Voice Clones, Professional Voice Clones, Voice Design voices, and Default voices — but not with Voice Library voices.
What is prompt strength in Voice Remixing?
Prompt strength controls how aggressively the remix transformation is applied. Low maintains most original characteristics with subtle changes. Medium offers balanced transformation. High strongly adheres to the remix prompt. Max produces a full transformation that may significantly change the voice’s identity. Start with Medium for most use cases.
Can I remix a Voice Library voice?
No — Voice Remixing only works with voices that belong to your account (your own clones, Voice Design voices, and Default voices). To modify a Library voice, first create a Voice Design voice with similar characteristics, then apply remixing to your owned version.
Does Voice Remixing change the original voice?
No — remixing creates new voice versions rather than overwriting the source. Your original voice is preserved. Each remix result is saved as a new voice in your library, maintaining the full lineage of the original source voice.
How is Voice Remixing different from Voice Design?
Voice Design creates entirely new synthetic voices from text prompts — no existing voice required. Voice Remixing modifies an existing voice you own. Use Voice Design when starting from scratch. Use Voice Remixing when you have an existing voice that needs modification.
Methodology
Voice Remixing capabilities from ElevenLabs official Voice Remixing documentation at elevenlabs.io/docs/overview/capabilities/voice-remixing. API parameters from ElevenLabs API reference at elevenlabs.io/docs/api-reference/text-to-voice/remix. LinkedIn launch announcement from ElevenLabs official LinkedIn post. Default voice compatibility from ElevenLabs text-to-speech product documentation. This article was drafted with AI assistance and reviewed by the editorial team at ElevenLabsMagazine.com.
References
ElevenLabs. (2026). Voice Remixing documentation. https://elevenlabs.io/docs/overview/capabilities/voice-remixing
ElevenLabs. (2026). Remix a voice API reference. https://elevenlabs.io/docs/api-reference/text-to-voice/remix
ElevenLabs. (2026). Text to Speech product guide. https://elevenlabs.io/docs/eleven-creative/playground/text-to-speech
ElevenLabs. (2026). Voice Remixing quickstart. https://elevenlabs.io/docs/eleven-api/guides/cookbooks/voices/remix-a-voice
