ElevenLabs Voice Cloning Not Working? Fix Verification, Training and Voice Errors

Start with the symptom, not a new subscription

When an ElevenLabs voice clone is not working, first separate four different problems: identity verification did not pass; a Professional Voice Clone is still being trained; the finished voice sounds wrong; or an audio upload does not meet the input requirements. The right solution depends on which of these is actually happening.

The diagnostic below uses ElevenLabs' published guidance, checked on November 3, 2026. Claim26 has not accessed a customer's workspace or performed a paid voice-cloning benchmark. Interface names, available voice models and plan permissions may change, so confirm the status displayed in your own account before deleting anything.

Fix Professional Voice Clone verification failures

Professional Voice Cloning (PVC) verifies that you are using your own voice. If the verification recording fails, begin with browser microphone access: make sure the correct microphone is selected, the browser is permitted to record and no hardware or software mute is enabled. Record in a quiet room and use a speaking style similar to your training samples.

ElevenLabs specifically advises reading each prompted verification line once, then pressing Stop. Repeating the same line can cause failure. If you have exhausted your verification attempts, the official help page says you can try again after 24 hours, or contact ElevenLabs support to review whether the attempts can be reset. Do not try to circumvent identity checks or verify somebody else's voice on your account.

What 'No model found for this voice' really means

The message 'No model found for this voice. Please select another voice' is often a timing problem, not proof that your clone has been permanently rejected. ElevenLabs says the message can appear when a Professional Voice Clone is selected before its dedicated voice model has finished fine-tuning and become available.

Open My Voices, find your Professional Voice Clone and view its model status. ElevenLabs says fine-tuning commonly takes about three to six hours and may take as long as 24 hours, depending on the queue. The company sends a notification when processing finishes. Avoid deleting a clone solely because you saw this message soon after creating it.

When Professional Voice Clone training fails or is delayed

A delayed voice model and a failed training run are not identical. In My Voices, check the model's status rather than assuming a stalled screen means that the voice data has been lost. ElevenLabs documents an automatic-retry status: in that case the system has already queued another attempt and says no immediate action is required from you.

If repeated training failures persist, the source recordings may need review. ElevenLabs notes that deleting and recreating a voice with improved data has helped some users, but this is a more disruptive step: save the original recordings, note the exact displayed message and consider support before starting over. Do not claim a retry will certainly succeed; quality problems can remain in the training set.

Fix an Instant Voice Clone that does not sound like you

Instant Voice Cloning (IVC) makes a quick approximation from relatively little speech. The official documentation recommends around one to two minutes of high-quality audio. It is not the same process as training a dedicated Professional Voice Clone, and an unusual accent or distinctive vocal character can be difficult to reproduce from a brief sample.

Before paying for a different tier, replace noisy or mismatched recordings. Use one speaker, clear speech, a consistent room and microphone distance, natural delivery and minimal reverb. Create a small comparison with the same script and model settings, then listen for tone, rhythm, consonants and accent rather than judging only the first few seconds. A better source sample may help more than a longer one.

Why the accent, pronunciation or emotional tone sounds wrong

ElevenLabs cautions that you cannot simply change the accent or tone of an existing clone after it is created. The source material strongly influences the resulting speaker style. A calm studio reading, highly expressive acting and a telephone recording are different kinds of training input; combining them without a plan can make the outcome unpredictable.

Check the language and the model selected for generation, then compare a short neutral passage and a passage containing names, dates and unfamiliar words. A pronunciation dictionary may help with a particular name, but it does not fix a clone trained on poor recordings. If Instant Voice Cloning cannot represent your accent reliably, Professional Voice Cloning may be worth considering if you own the voice and have enough clean audio.

Audio upload rejected? Check these requirements first

For Instant Voice Cloning, ElevenLabs' help documentation says each uploaded sample must last at least five seconds and the combined samples must reach at least ten seconds. That is an upload minimum, not its recommendation for good results: the recommended amount is approximately one to two minutes of good speech. Very short clips can pass validation and still sound unconvincing.

For file quality, ElevenLabs recommends MP3 at 192 kbps or higher, along with speech free from music, background conversations, echo and long pauses. WAV is not automatically a better choice and can sometimes create upload problems. Re-export a clean representative segment if the file is damaged or incorrectly encoded; do not inflate an otherwise unusable recording merely to meet a duration threshold.

Professional cloning requires a different recording plan

Professional Voice Cloning requires materially more clean audio than Instant cloning: ElevenLabs recommends approximately 30 to 180 minutes, with consistent delivery and one clearly identifiable speaker. Use several passages with the speaking style you actually want to reproduce. Adding hours of noisy or inconsistent recordings is not an effective substitute for a coherent dataset.

Professional cloning is available on eligible paid plans, starting at Creator in the provider's documentation, and verification is tied to the owner of the voice. Record your own voice, keep the originals and confirm the current plan's voice slots before upgrading. If the goal is to use somebody else's narration, follow ElevenLabs' documented permission and sharing process instead of attempting PVC verification as that person.

Use a fair listening test before replacing the clone

A practical comparison should hold the script, generation model, language and approximate voice settings constant. Choose three sample passages: an ordinary explanation, a passage with difficult names and a more expressive line. Save each output and log which issues recur, including pronunciation drift, accent mismatch, unstable pauses and unnatural breath sounds.

This is a suggested workflow rather than a claim that Claim26 tested actual voices. It helps you determine whether the problem belongs to the clone, the generation model or a particular script. The newest model is not automatically better for every recording; changing several settings at once prevents you from seeing which change helped.

Know when to wait, recreate or ask support

Wait and check status when a Professional Voice Clone is newly created and still training. Retry verification only according to the published verification procedure. Recreate a clone when the recordings are demonstrably unsuitable, or after persistent training issues and consideration of the original data. Contact ElevenLabs support for repeated verification failures, recurring training errors or an account-specific block that the published steps do not resolve.

Before submitting a support request, collect the exact error text, whether the clone is Instant or Professional, the current status shown in My Voices, approximate time since creation and a description of the microphone or upload format. Avoid sending private raw recordings through public forums or sharing account credentials. A support request should give enough context without exposing personal voice data unnecessarily.

Does your subscription or licence affect the fix?

A technical fix and permission to use a clone commercially are separate questions. ElevenLabs documents Instant Voice Cloning on eligible paid plans such as Starter, while Professional Voice Cloning requires Creator or above. Features and quotas can change, so check the live pricing page and the number of available voice slots before choosing a subscription.

The provider's legal guidance distinguishes Free-plan output from commercial rights on eligible paid plans and imposes conditions on generated content. Payment does not establish ownership of another person's identity or recording. If you are preparing a video, course or customer project, verify the voice permissions, source-script rights and active plan terms before publication. Claim26 may receive a commission through its disclosed ElevenLabs affiliate links, at no additional cost to you.

Frequently asked questions and next steps

Can 'No model found' be fixed by switching to another voice model? If your PVC is still training, the documented answer is to check its model-specific status and wait until the relevant model is ready. Repeatedly selecting another model does not substitute for completed processing. If the issue continues after the stated processing window, consult support and provide the status displayed in My Voices.

Should you upload more audio to an Instant clone that sounds bad? Not automatically. ElevenLabs says cleaner, consistent audio matters more than accumulating minutes, and more than two or three minutes may offer little improvement. If you need the full recording and ownership workflow, follow Claim26's separate voice-cloning setup guide; this page focuses on troubleshooting after something goes wrong.

Can you verify a Professional Voice Clone with a consenting actor's recordings? No: ElevenLabs says PVC is for the account holder's own verified voice. An actor can create and verify their own PVC and share it using permitted product features. If your problem is actually choosing between an original synthetic voice and a recognizable real voice, the separate Voice Design versus Voice Cloning comparison explains the trade-off.