ElevenLabs

Editorial assessment

Size the plan by finished minutes of narration, then add headroom for regenerations and for anything else drawn from the same pool. The pricing table equates each plan’s text-to-speech credits with roughly one minute of speech per 1,000 credits.

• Free

  for trying voices and for noncommercial projects under about 10 minutes of speech a month, published with attribution.

• StarterUSD 5 per month

  once audio is used commercially or you need instant voice cloning, up to about 30 minutes a month.

• CreatorUSD 18.33 per month

  for regular narration up to about two hours a month, or when you need a professional voice clone.

• ProUSD 82.50 per month

  for about 10 hours a month (600,000 credits), or when an API workflow needs 44.1kHz PCM output.

• ScaleUSD 249.17 per month

  ($299 a month, 1.8M credits) when three people need shared workspace seats.

• BusinessUSD 825 per month

  for about 100 hours of speech a month (6M credits), or when up to ten people and ten professional voice clones share one workspace; Enterprise once you need custom SSO, contract SLAs or a HIPAA BAA.

Transcription and especially dubbing use credits far faster than narration, so a dubbing-heavy project can need a tier or two above what its narration alone suggests. Annual billing charges ten months for a year, so Pro drops to $82.50 a month from $99. If the picture matters more than the sound, see the Runway profile; the audio and voice category lists other voice options.

On a paid plan, unused credits roll over for up to two months, so the balance can reach three times the monthly quota. Rollover does not apply to Free, and downgrading or cancelling forfeits unused credits at the end of the billing cycle; upgrading carries the unused quota into the new plan.

Pay As You Go credits are separate: you can prepay without a subscription or top up when the monthly credits run out. They are stored as a dollar value and converted at your current plan’s rate, so changing plans can change how many credits that balance buys.

No. Rights follow the plan in effect when the audio was generated. Free-generated audio is for noncommercial use with attribution, and buying a paid plan later does not change it. Audio generated on a paid plan comes with commercial rights, subject to the terms, the rights to your source material and exclusions such as beta features. Keep the generation date and the license terms with each project.

Voice Isolator works on an existing recording: upload an audio file or record through your microphone, run isolation, then play or download the new audio. It accepts files up to 500 MB and one hour; both limits matter before uploading a long interview. Keep the original alongside the processed file so you can compare quieter words and breaths.

The feature page advertises 10 minutes on Free for personal use and lists commercial use from Starter. That makes Free suitable for a short personal cleanup check, while a client deliverable needs the applicable paid rights and permission for the source recording. The web allowance is not a separate pool on top of your other audio usage.

The provider says this model is not optimized for extracting sung vocals from music. For an interview, judge whether the words remain intact rather than treating a quieter background as success; a missing consonant can make the result less useful. Website and API pricing are documented separately, so check the applicable rate before automating a batch.

“Our Tuesday session starts at 9:45, runs for 2 hours, and costs 149 euros per person. Dr. Joaquín Ferreira leads it. Registration closes on the 3rd of March — after that, the waitlist opens.”

Download the cleaned recording and reopen it outside the app. Accept the cleanup only if every number and word remains audible, the complete clip plays, and the noise reduction helps; reject missing syllables or altered meaning. For the generated narration, check pronunciation and pauses separately. Stop if credits or download access are unavailable, and keep Free output noncommercial. You can put two voice tools side by side, and our editorial policy explains how profiles are checked and corrected.