Public-data comparison · Updated 2026-09-18

What AI voice cloning costs, and whose voice you may clone

Q: How much does AI voice cloning cost?

No vendor checked charges a separate fee for a voice clone. The cost is the subscription tier that unlocks it, and the tiers vary enormously. A fast instant clone is available on free or near-free plans at ElevenLabs, Cartesia Pro at $5 a month, Synthesia on every plan and Descript on its free tier. A high-quality professional clone is rationed by slot: ElevenLabs gives zero slots on Free and Starter, one on Creator at $22 a month, one on Pro at $99, three on Scale at $299 and ten on Business at $990. Cartesia gives none below Startup at $49 a month, which includes two, and four on Scale at $299. So the cheapest professional voice clone in this directory is an ElevenLabs Creator subscription at $22 a month, or $18.33 billed annually.

Q: Can you clone someone else's voice if you have their permission?

It depends on the product, and on one vendor the answer is no at any price. The ElevenLabs professional voice cloning documentation states verbatim: You can only create a Professional Voice Clone of your own voice. Even with their consent, you cannot clone someone else's voice. Synthesia takes the same line, stating that it only allows the cloning of your own voice. Neither restriction lifts on a higher plan. ElevenLabs instant voice cloning is not documented as carrying that self-only rule and asks you to confirm you have the right and consent. Speechify runs the opposite design: its API takes the speaker full name, issues a single-use phrase for that person to read aloud, and keeps the recording as the consent record, which is a flow built specifically for cloning a third party with proof. If you are an agency cloning a client voice, this distinction decides your vendor before price does.

Q: Which plans include professional voice cloning?

On ElevenLabs, Creator and above. Its documentation gives the allocation as no slots on Free and Starter, one slot on Creator and Pro, three on Scale and ten on Business. Note that Starter at $6 a month adds a commercial licence but still cannot make a professional clone at all. On Cartesia, Startup and above: the pricing page lists instant voice cloning on Pro at $5 a month with professional cloning unavailable, two professional clones on Startup at $49 and four on Scale at $299. A useful comparison falls out of that: at an identical $299 a month, Cartesia Scale includes four professional clones and ElevenLabs Scale includes three.

Q: How much audio do you need to clone a voice?

Between ten seconds and three hours, depending on which kind of clone you want. Cartesia states that 10 seconds of audio is enough to get started, and up to 60 seconds to better retain the speaker accent. Speechify asks for a 10 to 30 second sample under 5MB. Synthesia accepts an upload of 1 to 5 minutes. ElevenLabs instant cloning recommends approximately 1-2 minutes and warns against more than three minutes, saying it will yield little improvement and can in some cases even be detrimental to the clone. ElevenLabs professional cloning is the outlier: 30 minutes is the bare minimum, it recommends closer to 2-3 hours of audio, and training then takes roughly 3 to 6 hours. That is about 1,080 times more audio than Cartesia asks for.

Q: Can you clone a voice for free?

You can usually make one, but on some plans you cannot publish what it says. ElevenLabs documentation states that instant voice cloning is available to all users, Descript lists limited custom voice clones on its free plan, and Synthesia states that voice cloning is available on all plans. The catch is the licence rather than the feature. ElevenLabs terms restrict free users to non-commercial purposes, and its help centre states that content created outside a paid subscription cannot be used commercially whether it was made before or after you pay, so a free clone is an evaluation and the audio has to be regenerated once you subscribe. Synthesia lists downloading your videos and removing the logo as paid additions rather than giving free users a distributable file.

No vendor on this list sells a voice clone as a line item. What you buy is a slot and a permission — and the permission is the part that has no price at all. ElevenLabs states plainly that you may only clone your own voice at its professional tier, "Even with their consent, you cannot clone someone else's voice", while Speechify built an API endpoint whose entire purpose is cloning someone else with proof. On price, the cheapest professional clone here is $22 a month on ElevenLabs Creator — and every tier above it charges about $99 a month per slot, flat.

Five round pedestals on a dark teal surface, four topped with empty glass cylinders and the centre one holding a single glowing cyan orb

Voice tools by job Commercial use rights All AI voice tools

Quick verdict

Review type: Public-data comparison, not hands-on testing. CartSignal has cloned no voices. Every figure and quotation below is transcribed from the vendor's own pricing page, documentation or help centre, checked 18 September 2026. The per-slot arithmetic is ours and is labelled where it appears.

Short answer: Decide whose voice it is first. If it is yours, buy on price and ElevenLabs Creator wins. If it belongs to a client, a colleague or a hired actor, two of the vendors here will not do it at any price, and that eliminates them before you compare rates.

Not legal advice. This page records what vendors publish about their own rules. Consent, likeness and publicity law is separate, varies by country, and is not covered here.

Disclosure: ElevenLabs has an affiliate redirect on this site, disclosed on its profile. Cartesia, Synthesia, Speechify, HeyGen, Murf and Descript carry no affiliate relationship in the local CartSignal data, and the vendor links below are plain official URLs.

The gate is a slot, not a price

Search for what voice cloning costs and you get per-character rates, because that is what the tools publish. But the rate is what the clone says. The clone itself is not metered anywhere in this directory. It is rationed.

Two vendors ration it the same way, with the same vocabulary — a cheap or free instant clone, and a scarce professional one allocated in slots.

Plan	Monthly	Instant clone	Professional clones	Cost per professional slot
ElevenLabs Free	$0	Yes	0	Not available
ElevenLabs Starter	$6	Yes	0	Not available
ElevenLabs Creator	$22	Yes	1	$22.00
ElevenLabs Pro	$99	Yes	1	$99.00
ElevenLabs Scale	$299	Yes	3	$99.67
ElevenLabs Business	$990	Yes	10	$99.00
Cartesia Free	$0	No	0	Not available
Cartesia Pro	$5	Yes	0	Not available
Cartesia Startup	$49	Yes	2	$24.50
Cartesia Scale	$299	Yes	4	$74.75

Slot counts are first-party: the ElevenLabs professional voice cloning documentation gives them as "Free and Starter plan: No PVC slots available" and "Creator, Pro, and legacy Scale plan: 1 PVC slot", with its pricing page listing "3 Professional Voice Clones" on the current Scale tier and "10 Professional Voice Clones" on Business. Cartesia's counts are the figures on its pricing page. The cost-per-slot column is CartSignal arithmetic.

Above Creator, a slot costs a flat $99

Divide each ElevenLabs tier by the clones it includes and the ladder does something unusual: $99.00 at Pro, $99.67 at Scale, $99.00 at Business. Across three tiers spanning 45x in subscription price, the professional clone is priced within 0.7% of the same number. Billed annually the figures are $82.50, $83.06 and $82.50.

Creator sits outside that pattern at $22.00 a slot, which is 77.8% cheaper than the identical single slot on Pro. Moving Creator to Pro is 4.5x the price for exactly the same cloning allowance — you are buying credits, not clones. This is the same shape this site found in the Speechify API ladder and in ElevenLabs per-character pricing, where the entry tier holds the best unit rate and the discount stops early.

One consequence worth naming: ElevenLabs Starter at $6 a month is the tier that can license but cannot clone. Its pricing page adds "Commercial License" at Starter, yet the slot table gives it zero professional clones. The $16 step to Creator is what buys the first one.

Whose voice: the rule no money lifts

This is the finding that should change shortlists, and it is stated verbatim by the vendor. The ElevenLabs cloning documentation says:

"You can only create a Professional Voice Clone of your own voice. Even with their consent, you cannot clone someone else's voice."

It adds that "All Professional Voice Clones require a verification process to confirm that the voice belongs to you." There is no plan that removes this. Business at $990 a month buys ten slots, and all ten must be you.

Synthesia takes the same position on its voice cloning page: "Synthesia strictly prohibits cloning voices without explicit consent. We only allow the cloning of your own voice and require users to provide verbal consent during the voice cloning process."

Speechify is built the other way round. Its voice cloning guide describes a flow that only makes sense for a third party: "Speechify issues a single-use challenge phrase, the speaker reads it aloud, and you record them doing so." The API takes the speaker's full name as an input. It also states "There's no limit to the number of voices you can create" — no slots at all.

So the two designs are not different prices for one product. They are different products:

Self-only, rationed. ElevenLabs professional cloning and Synthesia. Best for a founder, narrator or creator cloning themselves.
Third-party, verified, unlimited. Speechify. The design an agency or localisation studio needs.

ElevenLabs instant cloning sits between them: its documentation carries no self-only rule and instead asks you to "confirm that you have the right and consent to clone the voice". So a consenting client's voice can be instantly cloned on ElevenLabs — just not professionally cloned.

Consent is implemented three different ways

Every vendor here requires consent. Only some of them can check it.

Vendor	Mechanism	Machine-checkable
ElevenLabs instant clone	Confirm you have the right and consent before saving	No — an attestation
ElevenLabs professional clone	Verification that the voice is yours; third parties barred outright	Yes, for you only
Synthesia	Speaker records consent using a randomly generated passcode	Yes
Speechify	API-issued single-use phrase, read aloud, retained as the record	Yes
HeyGen avatars	Consent video reading an on-screen script, under 30 seconds	Yes
Cartesia	No consent step documented in the cloning guide	Not documented

Synthesia's knowledge base is the most specific: "The voice speaker must record consent using a unique, randomly generated passcode." The recording must be spoken aloud — silent recordings are rejected — and kept under 60 seconds. HeyGen's consent video article covers avatars rather than voices and requires the script be read "exactly as written", adding that "The person in the consent video must be the same person featured in the avatar footage."

A passcode or an issued phrase is a real safeguard, because it cannot be recorded in advance or lifted from existing audio. A tickbox cannot.

New on 13 September: Speechify made it an API contract

Speechify's changelog entry dated 13 August 2026 replaces its old consent object with a recorded challenge. Its documentation gives the flow: "POST /v1/voices/consent-challenges with the speaker's full_name returns a phrase and an id. Show the phrase to the speaker exactly as it comes back, record them reading it aloud, and send that recording as consent_recording with consent_challenge_id on POST /v1/voices."

The timing matters if you are building now. The new shape is pinned to Speechify-Version: 2026-09-13 and, in Speechify's words, "it is callable now by pinning that version, and it becomes the default for new workspaces on that date" — which was five days ago. The old route is going: "The unverified flow is deprecated and will be switched off", on a window Speechify describes as deliberately shorter than its standard 12-month sunset. No switch-off date has been announced; Speechify says it will appear in the changelog and go to affected workspaces first, and that teams who cannot migrate in time should contact support for an extension. Existing clones are safe — "Existing cloned voices are unaffected and keep working, and synthesis endpoints are unchanged."

What you have to supply: ten seconds to three hours

Vendor and clone type	Audio required	Ready in
Cartesia instant clone	10 seconds, up to 60s for accent	Instant
Speechify	10-30 seconds, under 5MB	Not stated
ElevenLabs instant clone	1-2 minutes; avoid over 3	Near-instant
Synthesia	Upload of 1 to 5 minutes	Under 15 minutes
ElevenLabs professional clone	30 min minimum, 2-3 hours advised	3 to 6 hours

That is roughly 1,080x between the two ends. The quality ladder in voice cloning is paid for in recording time and patience, not in dollars — which is the opposite of nearly everything else this site has costed.

Cartesia is blunt about when the cheap option is the wrong one: "An instant voice clone works well for most voices. If the speaker has a rare accent, is a character voice, or carries distinctive qualities you need to preserve, use a Pro Voice Clone instead." ElevenLabs is equally blunt in the other direction on instant clones, warning that recording more than three minutes "will yield little improvement and can, in some cases, even be detrimental to the clone". More audio is not monotonically better.

One genuinely metered cloning cost does exist. Cartesia's pricing page lists 225 credits per added voice accent, so localising a cloned voice is billed per accent on top of the plan.

The free clone, and what it is worth

Cloning is unusually generous at the free end. ElevenLabs documentation states instant cloning "is available to all users". Descript's pricing page lists "Limited custom voice clones" on its free plan. Synthesia's knowledge base says "Voice cloning is available on all plans", noting only that some languages are Enterprise-only.

The restriction has moved to the licence instead, which is covered in full on commercial use rights for AI voice and video. In short: an ElevenLabs free clone cannot be used commercially, and because its licence attaches to the moment of generation rather than to the account, upgrading later does not license audio you already made. Synthesia sells the download and logo removal as paid additions rather than the generation itself. A free clone is an audition, not an asset.

Where the answer is not public

Four gaps are named rather than filled, because guessing them would be worse than leaving them open.

HeyGen's free tier. Two reads of HeyGen's pricing page today returned different answers — one showed a voice clone included on Free, the other showed the feature absent from Free entirely. Its paid tiers list "Voice Cloning" with no quantity anywhere. CartSignal publishes no HeyGen cloning count as a result.
Murf. Its voice cloning page names no plan and no price, and routes to a contact form offering "less than 30 minutes" or "more than 30 minutes" of source data. Reported elsewhere as a Business-tier or enterprise feature; that is a tracker claim, not a Murf one.
Cartesia's consent rule. Its cloning guide documents the audio requirements and the 16 MB file limit but no consent step at all, so whether third-party cloning is permitted is not established from that page.
Speechify's tier. Its docs say only that cloning "appears once your workspace is on a plan that includes voice cloning and has at least one API key", without naming which plans those are.

Worth noting where the market has moved: Resemble AI, for years a voice-cloning specialist, now prices deepfake detection on its pricing page and sells no voice cloning to new customers at all — while releasing its cloning models free under the MIT licence. The commercial value has drifted from making clones to catching them.

What to buy

Cloning yourself, on a budget: ElevenLabs Creator at $22 a month, or $18.33 annually, for the cheapest professional slot here. Do not step up to Pro for cloning — it is the same one slot.
Cloning yourself, several voices: Cartesia Startup at $49 for two slots, or Scale at $299 for four against ElevenLabs Scale's three at the same price.
Cloning clients or actors: Speechify, for the verified-consent flow and no slot cap. ElevenLabs instant cloning is the fallback if professional quality is not required.
Cloning inside a video tool: Synthesia, available on every plan with a passcode consent recording and a result in under 15 minutes.
Just testing: Clone free almost anywhere, but read the licence before you publish a word of it.

For what the cloned voice then costs to speak, see AI voice tools by job and the per-character ranking in ElevenLabs alternatives. For localising one, see what AI dubbing costs per minute.

Frequently asked questions

How much does AI voice cloning cost?

No vendor here charges a separate fee for the clone. You buy the tier that unlocks it. Instant clones are free or near-free — available to all users on ElevenLabs, $5 a month on Cartesia Pro, every plan on Synthesia, the free tier on Descript. Professional clones are rationed: zero slots on ElevenLabs Free and Starter, one on Creator at $22, one on Pro at $99, three on Scale at $299, ten on Business at $990; none below Cartesia Startup at $49, which gives two.

Can you clone someone else's voice if you have their permission?

Not on ElevenLabs professional cloning, which states that even with their consent you cannot clone someone else's voice, and not on Synthesia, which allows only your own. Speechify's API is designed for exactly this case, issuing a phrase for the speaker to read aloud as the consent record. ElevenLabs instant cloning asks you to confirm you have consent and does not carry the self-only rule.

Which plans include professional voice cloning?

ElevenLabs Creator and above; Cartesia Startup and above. At an identical $299 a month, Cartesia Scale includes four professional clones against three on ElevenLabs Scale.

How much audio do you need to clone a voice?

Ten seconds on Cartesia, 10-30 seconds on Speechify, 1-2 minutes for an ElevenLabs instant clone, 1-5 minutes on Synthesia, and 30 minutes minimum with 2-3 hours advised for an ElevenLabs professional clone, which then trains for 3 to 6 hours.

Can you clone a voice for free?

Usually yes, but you often cannot publish it. ElevenLabs restricts free users to non-commercial use, and its licence attaches to the moment of generation, so upgrading does not cover audio you already made. Synthesia charges for the download rather than the clone.