Free
Token calculator

AI voice cloning cost: plans, audio needed and consent

Cloning a voice costs from $5 a month (ElevenLabs Starter) or $16 (Descript Hobbyist), billed yearly. Plans, audio needed, consent steps and limits.

By · Updated

Quick comparison

ProviderFromStarting planFree tierOur rating
ElevenLabsChecked
$5/moStarter30k credits/month; commercial license; Instant Voice Cloning; 20 Studio projects; Dubbing Studio; Image & Video10k credits/month, non-commercial useNot ratedVisit
DescriptChecked
$16/moHobbyist10 media hours/month; 400 AI credits/month; 1080p watermark-free export; 1 person1 media hour/month, 720p with watermarkNot ratedVisit
Murf AIChecked
$19/moCreator24 hours of voice generation a year; 100 projects; commercial rights; no voice cloning10 minutes of voice generation in total, no downloads, no commercial rightsNot ratedVisit
WellSaidChecked
$10/moStarter240 minutes of downloaded audio a year; unlimited generation; English voices; commercial rights3 download minutes/month, no commercial rights, no card neededNot ratedVisit
SpeechifyChecked
$8.33/moStudio Starter86,400 Studio credits a year (voiceover uses 1 credit per second, so 24 hours); commercial rights; dubbingStudio: 600 credits, no commercial rights or MP3 downloads. Reader: 10 robotic voices up to 1.5x speedNot ratedVisit

The cheapest self-serve way to clone a voice is ElevenLabs Starter at $5 a month billed yearly ($6 monthly), which includes Instant Voice Cloning from 1-2 minutes of audio. A Professional clone, trained on 30 minutes to 3 hours of your own voice, starts on ElevenLabs Creator at $18.33 a month billed yearly ($22 monthly), and Descript includes custom voice clones from Hobbyist at $16 a month billed yearly. Murf offers cloning only on Enterprise, and WellSaid and Speechify Studio don’t offer self-serve cloning at all (prices as of October 2026).

We haven’t cloned a voice on any of these tools ourselves. Prices come from the site’s product data (checked October 17-18, 2026); requirements, consent steps and restrictions come from each vendor’s own docs, help center and terms, read October 25, 2026. This guide describes what the vendors require; it isn’t legal advice on using someone’s voice.

Which plans include voice cloning?

Tool and plan Price billed yearly Price billed monthly Cloning included
ElevenLabs Free $0 $0 None
ElevenLabs Starter $5/mo ($60/yr) $6/mo Instant Voice Cloning
ElevenLabs Creator $18.33/mo (about $220/yr) $22/mo Instant + 1 Professional clone
ElevenLabs Pro $82.50/mo ($990/yr) $99/mo Instant + 1 Professional clone
ElevenLabs Scale $249.17/mo (about $2,990/yr) $299/mo Instant + 3 Professional clones
ElevenLabs Business $825/mo ($9,900/yr) $990/mo Instant + 10 Professional clones
Descript Free $0 $0 Custom voice clones “Limited”
Descript Hobbyist $16/mo per person $24/mo Custom voice clones
Descript Creator $24/mo per person $35/mo Custom voice clones
Descript Business $50/mo per person $65/mo Custom voice clones
Murf Creator / Business $19 / $66 per month $29 / $99 None
Murf Enterprise Custom Custom Voice cloning (Enterprise only)
WellSaid Starter to Business $10 to $160/mo $19 to $49/mo Not offered to users
Speechify Studio Starter / Creator $100 / $300 per year Not listed New clones not available

How to read it:

  • ElevenLabs has two kinds of clone. Instant Voice Cloning (IVC) makes a voice from short samples “near instantaneously” and, ElevenLabs says, doesn’t train a custom model. Professional Voice Cloning (PVC) trains “a dedicated model on a larger set of voice data”. PVC starts on Creator, IVC on Starter; Free has neither.
  • Descript clones from its cheapest paid plan. Its pricing page lists “AI Speech with custom voice clones” on Hobbyist, Creator and Business.
  • Murf’s cloning is a sales process. Murf’s help center says voice cloning “is not a self-serve solution” and is “available only with our Enterprise plan”. Its API docs say the same of Instant Voice Cloning.
  • WellSaid doesn’t let users clone. WellSaid’s ethics page says it has “never allowed users to independently clone voices”.
  • Speechify Studio has paused new clones. Speechify’s Studio terms, updated October 9, 2026, say “Creating new cloned voices in Speechify Studio is not currently available”, with an exception for business customers who sign a separate agreement. Speechify’s developer API, a separate product, has its own cloning; its pricing isn’t covered here.

For the wider plan comparison, including how much voiceover each plan includes, see our ElevenLabs alternatives roundup and ElevenLabs vs Murf AI.

How much audio does each tool need?

Tool Clone type Audio the vendor asks for Ready in
ElevenLabs Instant 1-2 minutes recommended; at least 5 seconds per file and 10 seconds in total; avoid more than 3 minutes Near instantly
ElevenLabs Professional 30 minutes “bare minimum”; 2-3 hours for best results 3-24 hours (its docs give both 3-6 and 6-24)
Descript Custom voice clone A script of “about 90 seconds” plus a consent statement “Typically within minutes”
Murf (Enterprise, Studio) Custom voice About 1-2 hours of studio recordings per voice style, as 48 kHz, 16-bit WAV 1-4 weeks
Murf (Enterprise, API) Instant One sample of at most 30 seconds and 40 MB, sampled at 24 kHz or higher “Within a few minutes”

The details that matter:

  • ElevenLabs Instant: quality matters more than length. ElevenLabs asks for clear audio “without any reverb, artifacts, or background noise of any kind”, a consistent tone, and a level of -23 dB to -18 dB RMS with a true peak of -3 dB. Its FAQ recommends MP3 at 192 kbps or above.
  • ElevenLabs Professional: ElevenLabs’ own figures vary. One section gives 30 minutes as the minimum, its guide suggests at least an hour, and its FAQ gives 30 to 180 minutes. Samples must have a single speaker, in a supported language, and singing isn’t supported.
  • Descript: Descript’s help center doesn’t give a minimum length, only that you should “speak naturally, with varied tone and expression.” You can’t add more audio to an existing clone; to improve one, you make a new clone. Clones are based on US English pronunciation, so Descript warns accents may not carry over.
  • Murf: the two figures describe two different Enterprise routes. The help center’s 1-2 hours and 1-4 weeks are for a studio-built custom voice; the API’s 30-second sample is for instant cloning through Murf’s API.

This is where the tools differ most.

ElevenLabs

  • Instant clones: when you save the voice, you “confirm that you have the right and consent to clone the voice.” The Instant cloning docs describe no separate verification step.
  • Professional clones: “For now, we only allow you to clone your own voice.” ElevenLabs says that even with someone else’s consent, you can’t make a Professional clone of them. Before training, you verify the voice by reading lines aloud in the browser, using the same or similar microphone and tone as your samples. If verification fails, you wait 24 hours to retry or contact ElevenLabs support.
  • Use policy: ElevenLabs’ Prohibited Use Policy (updated October 9, 2026) bans cloning “the voice, identity, or likeness of any living individual without their explicit and informed consent or other valid authorization”, and cloning a deceased person “in a manner that is reasonably likely to cause harm.” It also says “You may not submit voice recordings of anyone under 18 to create or train a voice model.”

Descript

  • You record a consent statement in the app before Descript builds the clone. The statement must be read in English, even if you’ll generate speech in another language.
  • “Voice clones can only be created by the person themselves”, or by someone acting for them with their recorded authorization. A collaborator who can’t record in the app can send you a recording of the statement.
  • Descript won’t create clones of a deceased person, a non-consenting person, someone unable to record the statement, or audio from an AI source, and says “Any attempt to bypass this process is a breach of Descript’s Terms of Service.”

Murf

  • Murf’s Terms of Service (last updated December 19, 2024) define a “Consenting Speaker” as a person “who has given explicit written consent” to have their recordings used to train a voice. When you submit training audio, you represent that you have that written consent for any third-party speaker, and submitting a third party’s unauthorized recordings “is prohibited.”
  • Murf’s API docs add: “Only clone a voice you own or have clear, documented consent to clone.”

Speechify

  • Studio terms: “YOU MUST BE AT LEAST 18 TO USE ANY VOICE CLONING OR ORIGINAL VOICE DUBBING FEATURE.” Existing clones can be used only if the voice owner completed Speechify’s Voice Owner verification; “Voices created on the basis of a user’s attestation alone cannot be used.”
  • Speechify’s API docs describe a recorded consent step: the speaker reads a single-use challenge phrase aloud, and that recording is the consent record. The sample itself is 10-30 seconds and under 5 MB, and cloning “is not available in some jurisdictions,” based on the IP address of the machine making the request.

WellSaid

There’s no user consent step because users can’t clone. WellSaid says its voices are built from “private datasets from individuals who have given explicit and written consent”, and that its voice actors “are paid residuals for customer usage on our platform.”

What are you not allowed to do with a cloned voice?

Each vendor documents its own limits:

  • ElevenLabs bans impersonating political candidates or elected officials without authorization, and bans political campaigning without its prior written consent. You must “clearly disclose when people are interacting with AI”, and you must not remove or hide its watermarks or other provenance signals. Circumventing its voice verification systems, “such as Voice CAPTCHA”, is also banned.
  • Speechify Studio bars using a cloned voice of “a reasonably well-known current or former political figure or of a deceased individual”, impersonating anyone “for deceptive purposes”, and using a voice after the owner’s authorization has expired or been withdrawn. You may use a clone only “within the scope of that consent.”
  • Murf bars unauthorized recordings in training audio. Its help center says you own the rights to a custom voice and can use it commercially.

Is there free voice cloning?

Only in a limited form, among these five:

  • ElevenLabs Free includes 10,000 credits a month but no voice cloning; Instant cloning starts on Starter. The free plan is also non-commercial.
  • Descript Free includes 60 media minutes a month, 100 AI credits (one-time) and a “Limited” allowance of custom voice clones and text to speech, with watermarked 720p export. Descript’s voice cloning page calls it a “Limited trial of AI Speech.”
  • Murf Free, WellSaid’s trial and Speechify Studio Free include no cloning.

How much does one cloned voice cost per month?

The plan is the cost: none of these vendors lists a separate fee for creating or keeping a clone. Murf’s API docs say clone speech is billed at standard Falcon 2.1 rates, with no extra charge to create or keep a clone. What changes the bill is how much speech you generate with it.

  • ElevenLabs prices by credits. Its pricing page puts Starter’s 30,000 credits at about 30 minutes of text to speech a month and Creator’s 121,000 at about 121 minutes. On its Multilingual v2 model, one character of text is one credit. Unused credits roll over for up to two months, capped at twice the monthly quota (not on Free).
  • Descript charges AI Speech in AI credits (400 a month on Hobbyist, 800 on Creator). Its help center says longer text-to-speech output uses more credits but publishes no per-minute rate, and unused credits don’t roll over.
  • Murf Enterprise is custom-priced, so you need a quote.

Worked example: a podcaster cloning their own voice

Say you want to fix and add lines in your own voice, about 1 hour of generated speech a month, for a year.

ElevenLabs Starter (Instant clone): about 30 minutes a month is included, so an hour won’t fit. If 30 minutes is enough, it’s the cheapest paid cloning here: $5 × 12 = $60 a year billed yearly, or $6 × 12 = $72 billed monthly.

ElevenLabs Creator (Professional clone): about 121 minutes a month covers 60 minutes with room to spare, and includes one Professional clone.

  • Billed yearly: $18.33 × 12 = $219.96 a year (ElevenLabs shows about $220)
  • Billed monthly: $22 × 12 = $264 a year
  • Yearly billing saves $264 − $219.96 = $44.04

You’d also need 30 minutes to 3 hours of clean recordings of yourself, and the voice must be yours.

Descript Hobbyist: $16 × 12 = $192 a year billed yearly, or $24 × 12 = $288 billed monthly. You also get the editor and 10 media hours a month. Whether 400 AI credits cover an hour of AI Speech a month depends on Descript’s credit rates, which it doesn’t publish.

What to watch for

  • Instant vs Professional isn’t just quality. On ElevenLabs, an Instant clone needs a consent confirmation; a Professional clone must be your own voice and passes a spoken verification. If you need to clone a colleague or a voice actor, ElevenLabs’ Professional route won’t allow it.
  • Clones are fixed. Descript and Murf’s API both say you can’t add training data to an existing clone; you make a new one.
  • Downgrading can lock a clone. On ElevenLabs, dropping below Creator leaves a Professional clone unusable until you upgrade.
  • Clones stay in the vendor’s account. ElevenLabs says you “cannot download or export your voice clones as standalone files,” so switching tools means cloning again.
  • Promotions expire. ElevenLabs’ pricing page currently shows first-month discounts on monthly Starter and Creator; the regular rates above apply after that.
  • Policies change. Speechify Studio doesn’t currently create new clones, and ElevenLabs updated its Prohibited Use Policy on October 9, 2026. Read the current terms before a project that depends on a clone.

Which should you choose?

  • You want to try cloning your own voice cheaply: ElevenLabs Starter, with Instant cloning from 1-2 minutes of audio, at $5 a month billed yearly.
  • You want the most realistic clone of your own voice: ElevenLabs Creator and its Professional clone, if you can record at least 30 minutes of clean audio.
  • You edit podcasts or videos and want to patch lines in your voice: Descript Hobbyist. Its clone needs about 90 seconds of reading and a recorded consent statement, and it sits inside the editor.
  • A company needs a custom brand or presenter voice: Murf Enterprise (quote only), with written consent from the speaker and 1-2 hours of studio audio for its custom voice service.
  • You don’t need a clone, just good stock voices: WellSaid (whose voices come from consenting voice actors) or Speechify Studio. Compare them on our AI voice hub.

Sources