will V3 work with professional voices or only with designed voices? | ElevenLabs | Page 1

marble runeBOT Apr 9, 2026, 2:29 PM

#

AI-generated Answer for sturoro

Only designed voices. Per the docs, Eleven v3 supports voices created with Voice Design v3 and does not currently support Professional Voice Clones. PVCs are trained for and run on Flash v2.5, Turbo v2.5, and Multilingual v2 models, not v3. (help.elevenlabs.io)

Sources:
What is Voice Design? – ElevenLabs

cobalt fossil Apr 9, 2026, 2:32 PM

#

my professional voice gets punctuation pauses wrong. Often having no gaps with commas and full stop. What can be done?

marble runeBOT Apr 9, 2026, 2:32 PM

#

AI-generated Answer for sturoro

Short answer: punctuation alone isn’t a reliable pause control—especially with Eleven v3 and Professional Voice Clones (PVCs). Use explicit pause controls, adjust speed, and, if needed, tweak training data or model choice.

What you can do

If you’re on Eleven v3: insert audio tags directly in the text, e.g., [short pause], [pause], [long pause], at commas or sentence ends. In the UI you can auto‑insert tags with Enhance. Eleven v3 does not support SSML <break>. (help.elevenlabs.io)
If you’re on Flash v2.x/Turbo v2.x/English v1: use SSML <break time="x.xs" /> for precise pauses (up to 3 s). Avoid overusing breaks to prevent artifacts. Dashes (—) or ellipses (…) can help, but are less consistent. (help.elevenlabs.io)
Slow the delivery: lower the voice speed below 1.0 (range 0.7–1.2) via voice settings/API to lengthen natural gaps. (elevenlabs.io)
Mind the training data (PVC): pacing mirrors your samples. Use longer, continuous recordings and include natural sentence/paragraph pauses; add more material if needed. (elevenlabs.io)
If v3 + PVC still misbehaves: note PVCs are not yet fully optimized for v3; for v3 features consider a designed/IVC voice, or use Flash/Turbo for steadier delivery. (elevenlabs.io)
Try a regeneration; it often resolves quality issues. (elevenlabs.io)

If you tell me which model you’re using (v3 vs Flash/Turbo) and where you generate (Studio vs API), I can show the exact text/tag or payload pattern to fix the pauses. (elevenlabs.io)

Sources:
Do pauses and SSML phoneme tags work with the API? – ElevenLabs
How can I add pauses? – ElevenLabs
Get default voice settings | ElevenLabs Documentation
Controls | ElevenLabs Documentation
Best practices | ElevenLabs Documentation
Text to Speech | ElevenLabs Documentation
Text to Speech | ElevenLabs Documentation

cobalt fossil Apr 9, 2026, 2:36 PM

#

im using multilanguage v2, through an api. It also often pauses mid sentence

marble runeBOT Apr 9, 2026, 2:36 PM

#

AI-generated Answer for sturoro

Thanks for the detail. Mid‑sentence pauses with eleven_multilingual_v2 are usually caused by one (or a mix) of chunking without context, aggressive style settings, ambiguous language normalization, or punctuation that implies hesitation.

What to try

If you split or stream text, provide context so prosody carries across chunks:
- Use previous_text/next_text or previous_request_ids/next_request_ids when making multiple TTS calls to avoid pauses at chunk boundaries. (elevenlabs.io)
Lock the language and normalization:
- Set language_code (ISO 639‑1) so the normalizer uses the right rules; keep apply_text_normalization on auto (or on if you need numbers expanded consistently). (elevenlabs.io)
Tame style settings that can introduce instability:
- Keep style at 0; high style exaggeration can cause “inconsistent speed” and other artifacts. Adjust stability/similarity as needed for your voice. (elevenlabs.io)
Clean up punctuation and only add deliberate pauses:
- Ellipses (…) and dashes (—) can imply hesitation; for exact pauses, use <break time="1.0s" /> sparingly (up to 3s). Overusing break tags can itself cause speedups/artifacts. (help.elevenlabs.io)
Avoid aggressive latency optimizations:
- If you’re using optimize_streaming_latency, reduce or remove it; levels 3–4 trade quality (and 4 disables the text normalizer) and can hurt prosody. (elevenlabs.io)

Minimal example (HTTP TTS) showing the relevant fields

POST /v1/text-to-speech/:voice_id
Body:
- model_id: "eleven_multilingual_v2"
- language_code: e.g., "es" (or your target)
- voice_settings: { style: 0, stability: <your value>, similarity_boost: <your value>, speed: 1.0 }
- previous_text / next_text when chunking
  Reference: API parameters and usage. (elevenlabs.io)

If the issue persists across voices after the above, follow the troubleshooting guidance (style at 0, adjust stability/similarity) and contact support as advised there. (elevenlabs.io)

Related docs

Create speech API (all parameters above). (elevenlabs.io)
Best practices for pacing/speed ranges and handling inconsistent pauses.

cobalt fossil Apr 9, 2026, 2:40 PM

#

can i make a designed voice based on my professional voice?

marble runeBOT Apr 9, 2026, 2:40 PM

#

AI-generated Answer for sturoro

Short answer: not directly. Voice Design creates voices from a text prompt and doesn’t accept an existing voice (including a Professional Voice Clone) as a seed. (help.elevenlabs.io)

What you can do instead

Use Voice Remixing (alpha) to generate a new v3-compatible voice variant derived from your Professional Voice Clone. It modifies voices you own and outputs a new voice; your original PVC remains unchanged. (elevenlabs.io)
Via API: POST /v1/text-to-voice/:voice_id/remix with a voice_description to get previews (each has a generated_voice_id). Then create the new voice from the chosen preview. (elevenlabs.io)
Constraints: you can only remix voices you own (e.g., your PVC/IVC/designed/default); you cannot remix voices from the Voice Library. (elevenlabs.io)
Note: Eleven v3 currently does not support Professional Voice Clones directly; remixing into a v3 voice is a way to use v3 capabilities based on your PVC. (help.elevenlabs.io)

If you want, I can outline the exact API calls you’d make with your PVC’s voice_id to produce a remixed voice. (elevenlabs.io)

Sources:
What is Voice Design? – ElevenLabs
Voice remixing | ElevenLabs Documentation
Remix a voice | ElevenLabs Documentation
Text to Speech (product guide) | ElevenLabs Documentation
What is Voice Design? – ElevenLabs
Remix a voice | ElevenLabs Documentation

cobalt fossil Apr 9, 2026, 2:49 PM

#

when crea professional voice is it better to use one long audio example or several smaller samples?

marble runeBOT Apr 9, 2026, 2:49 PM

#

AI-generated Answer for sturoro

Short answer: the number of files doesn’t matter—total clean runtime does. For multi‑hour datasets, ElevenLabs recommends splitting into multiple ~30‑minute samples purely for easier uploading. Minimum effective length is ~30 minutes; for best quality, target 2–3 hours of consistent, single‑speaker audio. (elevenlabs.io)

Sources:
Professional Voice Cloning | ElevenLabs Documentation

cobalt fossil Apr 9, 2026, 2:51 PM

#

where do i find the voice remix?

marble runeBOT Apr 9, 2026, 2:51 PM

#

AI-generated Answer for sturoro

In the API: use POST /v1/text-to-voice/:voice_id/remix to generate remix previews, then create a new voice from the chosen preview’s generated_voice_id. Quickstart and parameters are here: Remix API reference and Voice Remixing quickstart. (elevenlabs.io)
In the UI (core platform): ElevenLabs states Voice Remixing is available on the core platform (currently in alpha). The Text to Speech guide references the “Voice Remixing” tool. (elevenlabs.io)

Notes/eligibility:

You can remix voices you created (Voice Design, IVC, PVC); you cannot remix voices from the public Voice Library. The TTS guide also notes default voices can be remixed. (elevenlabs.io)

Sources:
Remix a voice | ElevenLabs Documentation
Voice remixing | ElevenLabs Documentation
Voice Remixing quickstart | ElevenLabs Documentation

cobalt fossil Apr 9, 2026, 3:00 PM

#

what's the url for remix?

marble runeBOT Apr 9, 2026, 3:00 PM

#

AI-generated Answer for sturoro

API endpoint: https://api.elevenlabs.io/v1/text-to-voice/:voice_id/remix. (elevenlabs.io)
Notes: ElevenLabs’ public docs describe Voice Remixing and its API, but do not publish a direct in‑app Remix URL. See the feature overview for details. (elevenlabs.io)

Sources:
Remix a voice | ElevenLabs Documentation
Voice remixing | ElevenLabs Documentation

cobalt fossil Apr 9, 2026, 3:04 PM

#

the documentation says "ElevenLabs voice remixing is available on the core platform and via API"

marble runeBOT Apr 9, 2026, 3:04 PM

#

AI-generated Answer for sturoro

Yes. The official docs confirm: “ElevenLabs voice remixing is available on the core platform and via API.” (elevenlabs.io)

API usage (summary):

Preview a remix: POST /v1/text-to-voice/:voice_id/remix → returns preview samples and a generated_voice_id. (elevenlabs.io)
Save the remixed voice: POST /v1/text-to-voice with generated_voice_id to create the new voice in your library. (elevenlabs.io)
Step‑by‑step quickstart is here (includes code). (elevenlabs.io)

Eligibility/limits:

You can only remix voices you own (e.g., your IVC, PVC, or Voice Design voices). Voices from the Voice Library cannot be remixed. (elevenlabs.io)

cobalt fossil Apr 9, 2026, 3:43 PM

#

im using remix, but my starting voice sounds nothing like me

#will V3 work with professional voices or only with designed voices?