#My professional voice clone keeps failing the fine-tuning.

1 messages · Page 1 of 1 (latest)

modern thistleBOT
#
AI-generated Answer for efistoffeles

Thanks for reaching out—here’s how to triage repeated Professional Voice Cloning (PVC) fine‑tuning failures based on our official guidance:

  1. Check the exact status in My Voices
  • If you see “We are sorry the training run experienced issues and has been retried,” the system will automatically retry. If you encounter multiple failures, it likely indicates a dataset issue. (help.elevenlabs.io)
  • Some models require you to manually click to start fine‑tuning; look for “Click to start fine‑tuning” next to the model name. (help.elevenlabs.io)
  1. If it keeps failing, re‑upload from scratch
  • Delete the voice and re‑upload the training data; this resolves issues for some users. If it still fails, review/change your training audio. (help.elevenlabs.io)
  1. Validate your training dataset against PVC requirements
  • Duration: Minimum ~30 minutes; 2–3 hours is recommended for best results. Total runtime matters more than the number of files. (elevenlabs.io)
  • Single speaker only; remove background noise, music, reverb; keep volume and delivery consistent. (elevenlabs.io)
  • Language: Use samples in a supported language; uploads in unsupported languages are rejected. (elevenlabs.io)
  • Use the uploader’s processing tools to denoise and separate speakers before submitting. (elevenlabs.io)
  1. Common signs you’re not ready to submit
  • Mixed speakers, room noise, or inconsistent style across samples often lead to training problems and lower‑quality results. Align the samples to the intended use (e.g., all audiobook‑style if that’s your goal). (elevenlabs.io)
  1. Still failing?
  • Contact Support via the help center form and include your voice ID, screenshots of the status, and short audio snippets illustrating the issue. (help.elevenlabs.io)

References

  • My professional voice clone failed or is delayed, what can I do? (help.elevenlabs.io)
  • What does the status of my Professional Voice Clone mean? (help.elevenlabs.io)
  • Professional Voice Cloning (requirements, audio quality, processing) (elevenlabs.io)
  • What files do you accept for voice cloning? (length/quality recommendations) (help.elevenlabs.io)
  • Use samples in supported languages (unsupported uploads are rejected) (elevenlabs.io)
vapid grotto
#

It tried automatically multiple times and failed constantly, now it just shows "fine-tuning failed. Click to retry" and when I click it says the same thing. My audio has multiple speakers but I used the processing tools to denoise it and separate the speakers.

modern thistleBOT
#
AI-generated Answer for efistoffeles

Thanks for the detail — repeated “fine‑tuning failed” almost always points to issues in the training dataset rather than a transient platform error.

What to do next (in order):

  1. Start fresh: Delete this PVC and re‑upload the training data from scratch. This resolves many repeated fine‑tuning failures. (help.elevenlabs.io)
  2. Curate the dataset to meet PVC requirements:
    • Only one speaker (you) per clip; remove any other voices. (help.elevenlabs.io)
    • No background noise, music, heavy reverb; keep volume and tone consistent; avoid long silences. (help.elevenlabs.io)
    • Provide sufficient clean speech: 30–180 minutes total is recommended for PVC. (help.elevenlabs.io)
    • Light denoising is fine if needed, but prioritize naturally clean recordings. (help.elevenlabs.io)
    • Use MP3 at 192 kbps+ (WAV won’t materially improve training vs good MP3). (help.elevenlabs.io)
  3. Re‑submit and monitor status in My Voices; hover over each model to see progress and messages. If you repeatedly see retries/failures, it indicates a dataset issue that the AI can’t resolve. (help.elevenlabs.io)
  4. If it still fails after a clean re‑upload, contact Support so they can investigate the training runs. (help.elevenlabs.io)

Reference articles:

vapid grotto
#

Okey and can I upload my audio in m4a format from clipchamp?