Short answer: you can’t—v3 doesn’t support Professional Voice Clones yet.
Details
As of March 27, 2026, the Eleven v3 model “does not currently support Professional Voice Clones,” so an existing PVC (even from 2024) cannot be trained for v3. (help.elevenlabs.io)
Your PVC will instead auto-train on the supported models (Flash v2.5, Turbo v2.5, Multilingual v2), and you can add any newly supported fine-tuning models from My Voices when ElevenLabs enables them. You’ll see an exclamation icon next to the voice; hover it to view the notice, then click the plus icon next to the new model to start fine-tuning. (help.elevenlabs.io)
Workaround if you need v3 today
Create a Voice Design v3 voice; Voice Design v3 voices are fully compatible with the Eleven v3 model and its advanced controls. (help.elevenlabs.io)
Keep an eye on My Voices for future support notices; once v3 (or another new model) supports PVC fine-tuning, you’ll be able to trigger it from there. (help.elevenlabs.io)
Remixing creates a new voice variant whose output is a full‑quality v3 voice model (backwards compatible with other models). Your original voice remains unchanged. (elevenlabs.io)
You can remix from a Professional Voice Clone to produce this v3-compatible variant. (elevenlabs.io)
Note: Eleven v3 does not directly support Professional Voice Clones; remixing is the path to use a PVC with v3. (help.elevenlabs.io)
If you’re using the API, the Remix endpoint returns previews with a generated_voice_id; create/use the remixed voice and synthesize with model_id "eleven_v3" on the speech endpoints. (elevenlabs.io)
Short answer: There’s no published ETA. As of March 27, 2026, Eleven v3 does not currently support Professional Voice Clones (PVCs). (help.elevenlabs.io)
Current state and what to expect:
PVCs train on Flash v2.5, Turbo v2.5 and Multilingual v2; if new models add fine-tuning support you’ll see a notification in My Voices. (help.elevenlabs.io)
Eleven’s v3 launch post notes PVCs are “not fully optimized for Eleven v3” and that optimization is “coming in the near future,” but gives no date. (elevenlabs.io)
Workarounds if you need v3 features now:
Use Voice Design v3 (fully compatible with Eleven v3). (help.elevenlabs.io)
Or use Voice Remixing: you can input a voice you own (including a PVC) and output a full-quality v3 voice model. (elevenlabs.io)
For clarity on where PVCs work today, see the PVC language/model support article. (help.elevenlabs.io)
lilac slate
high anchor