# 6 best voice cloning tools from 10 seconds of audio in 2026

By Samuel Oram-Jones

ElevenLabs, Cartesia, MiniMax, Gradium, and Bland Speech all publish 10 seconds as the length for an instant clone, so the number alone no longer separates them. What separates them is what happens around that sample: whether a clone needs a $0 plan or a $6 one, whether the provider checks the speaker, and what the voice sounds like once it is speaking. Bland puts the limit plainly, saying the quality of the sample sets the ceiling on the quality of the voice.

Ten seconds also isn't the floor. Inworld clones from 3 seconds, Typecast from 5, and Smallest AI from 5 to 15.

## What is the best voice cloning from 10 seconds of audio?

Bland Speech is the best voice cloning from 10 seconds of audio for a clone that will speak on a phone call, as of November 2026, because one clean sample of about 10 seconds makes an instant clone that Bland says is ready in a minute, and the model that speaks it, Bland Speech v3, ranked first of 13 AI voices at 1365 in Design Arena's [Audio Realism Bench results of 4 August 2026](https://notes.designarena.ai/audio-realism-benchmark-measuring-realism-in-audio-models/). Bland Speech doesn't ask for the least audio, since Inworld asks for 3 seconds, so this list puts it first on what the clone speaks through rather than on sample length.

Design Arena scored each model's own stock voices, not clones, so the result says how Bland's voices sound and nothing about how closely a clone matches its speaker.

- **Bland Speech:** Best overall for a 10-second clone that will speak on phone calls
- **ElevenLabs:** Best for a 10-second clone on a $6 plan with a verification step
- **MiniMax:** Best for a 10-second clone billed at $1.5 a voice
- **Speechify:** Best for a 10 to 30 second clone with a spoken consent check
- **Typecast:** Best for a clone from as little as 5 seconds through an API
- **Gradium:** Best for a 10-second clone on a real-time voice agent API

| Service | Audio for an instant clone | Plan or price to make one | Sample rules | Speech price per million |
|---|---|---|---|---|
| Bland Speech | About 10 seconds | Pay as you go, no monthly plan | One file up to 10 MB | $40 |
| ElevenLabs | 10 seconds on Eleven v4 | Starter at $6 a month | Verification step on the clone | $80 or $40, by model |
| MiniMax | 10 seconds to 5 minutes | $1.5 per voice, charged at first use | mp3, m4a, or wav up to 20 MB | $100 or $60, by model |
| Speechify | 10 to 30 seconds | Starter at $10 a month | Speaker reads a challenge phrase | $15 on Starter for Simba 3.5 |
| Typecast | 5 to 150 seconds on the API | API Lite at $15 a month | WAV or MP3 up to 25 MiB | $75 on Lite, at 1 credit per character |
| Gradium | 10 seconds recommended | Free plan, with 6 custom voices | Clear audio, one speaker | $36 to $58 per million credits, by plan |

<details><summary>What these rows mean</summary>

- **Audio for an instant clone:** the sample length each company documents for a clone made from one recording with no training run.
- **Plan or price to make one:** the cheapest plan or charge that the company ties to making a clone.
- **Speech price per million:** each company's published rate for speech. Typecast and Gradium bill 1 credit per character.

</details>

A buyer who needs the least possible audio should look at Inworld first, and a buyer testing on a $0 plan can start on Gradium's Free plan.

## What does a 10-second sample actually buy?

A 10-second sample buys a recognizable voice that is only as clean as the recording, and the providers agree on that more than they agree on anything else here. Bland asks for one speaker, no background noise, music, or reverb, at least 100 ms of silence at each end so no word is clipped, and some emotional variety, which yields a more expressive clone. A phone recording with a fan running makes a clone with a fan in it.

A longer sample helps most providers, and the gain differs by company. Cartesia says 10 seconds is enough to get started and up to 60 seconds better retains the speaker's accent on Sonic 3.6. Inworld caps a sample at 30 seconds and says speaker similarity improves with length up to that cap. Fish Audio asks for at least 10 seconds per clip and says a minute or two of clear speech improves fidelity. Bland's V3 clone engine takes exactly one file, with roughly 10 seconds ideal, so a cleaner take matters more there than a longer one.

## How do the 6 best voice cloning tools from 10 seconds compare?

These six differ most in what a clone costs to start and what a clone speaks through.

### 1. Bland Speech: Best overall for a 10-second clone that will speak on phone calls

**What does it do?** Bland Speech makes an instant clone from one clean sample of about 10 seconds, in a single file of up to 10 MB, and a professional clone from 30 minutes or more of audio after a consent and rights review. Each clone is private to the customer's organization, the customer confirms every voice is theirs to use, and the clone's voice ID goes into Bland's speech calls. Output arrives as PCM from 8 to 48 kHz or 8 kHz mu-law, and Bland publishes 199ms to first audio.

**Who is it for?** Bland Speech suits a team that wants a voice from one short recording speaking on phone calls and voice agents.

<div class="cmp-proscons cmp-proscons-own"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>An instant clone from one sample of about 10 seconds</li><li>Clones private to a customer's organization</li><li>Bland Speech v3 ranked first of 13 AI voices on Audio Realism Bench on 4 August 2026</li><li>Mu-law and PCM output ready for phone calls</li><li>Pay as you go with no monthly plan</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>More than a one-off narration project needs, because Bland Speech is built for production calls</li></ul></div></div>

**Pricing:** $0.04 per 1,000 characters, which is $40 per million, pay as you go with no monthly plan, and 50,000 characters free once for a new account. The [Bland Speech](/speech) page has a studio to try a clone.

### 2. ElevenLabs: Best for a 10-second clone on a $6 plan with a verification step

**What does it do?** ElevenLabs says Instant Voice Clones on Eleven v4 capture voices with high fidelity from just 10 seconds of audio, and sells instant cloning from its Starter plan at $6 a month with a commercial license. Professional cloning needs about 30 minutes of audio and the Creator plan at $22 a month. Instant clones carry a verification step that ElevenLabs describes as an ethical and legal safeguard.

**Who is it for?** ElevenLabs suits a team that wants to make a clone from a dashboard under a monthly plan.

<div class="cmp-proscons" data-rival="elevenlabs"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Instant clones from 10 seconds of audio on Eleven v4</li><li>A commercial license from the $6 Starter plan</li><li>Mu-law and A-law output at 8 kHz</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Instant cloning needs a paid plan, and the Free plan is non-commercial</li><li>Eleven v3 ranked 12th of 13 AI voices on Audio Realism Bench on 4 August 2026, and Eleven v4 came out on 28 September 2026, after those results</li><li>The standard text to speech WebSocket doesn't accept Eleven v3 or v4</li></ul></div></div>

**Pricing:** $0.08 per 1,000 characters for Eleven v4, v3, and Multilingual v2, and $0.04 for v4 Turbo, v3 Conversational, and Flash, against plans from Free to Business at $990 a month.

### 3. MiniMax: Best for a 10-second clone billed at $1.5 a voice

**What does it do?** MiniMax's Rapid Voice Cloning takes a recording of 10 seconds to 5 minutes, as mp3, m4a, or wav up to 20 MB, and an optional prompt recording under 8 seconds improves similarity. MiniMax charges $1.5 per voice the first time the voice is used for speech, and deletes a cloned voice that isn't used within 7 days.

**Who is it for?** MiniMax suits a developer who wants to pay per clone and will keep each clone in regular use.

<div class="cmp-proscons" data-rival="minimax"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>$1.5 per cloned voice, charged at first use</li><li>A bidirectional WebSocket that takes language model output token by token, with a cancel message</li><li>Mu-law output at 8 kHz</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>A cloned voice is deleted after 7 days without use</li><li>Speech costs $60 to $100 per million characters</li><li>Speech-02 HD ranked 7th of 13 AI voices on Audio Realism Bench on 4 August 2026, and Speech 2.8 isn't in those results</li></ul></div></div>

**Pricing:** $100 per million characters for speech-2.8-hd and $60 for speech-2.8-turbo on pay as you go, plus $1.5 per cloned voice, or monthly Audio Packages from $5 for 100,000 audio points, with 10 voice slots on Starter.

### 4. Speechify: Best for a 10 to 30 second clone with a spoken consent check

**What does it do?** SpeechifyAI clones a voice from a 10 to 30 second sample with verified consent. The speaker reads a challenge phrase, Speechify checks it and retains it as the consent record, and no path skips the step. A clone works in all 48 languages on Simba 3.5 and keeps the accent of the recording.

**Who is it for?** Speechify suits a team that wants a consent recording stored for every clone and will pay from $10 a month.

<div class="cmp-proscons" data-rival="speechify"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Consent verified by a spoken challenge phrase</li><li>No limit on the number of voices a customer can create</li><li>Generated audio carries a watermark, with detect and verify endpoints</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Cloning is on paid plans, with none on Free</li><li>Cloning is refused from some jurisdictions</li><li>Only models redeployed since the watermark shipped mark their output</li></ul></div></div>

**Pricing:** Starter at $10 a month, Pro at $99, and Scale at $499, with Simba 3.5 at $15, $12, and $10 per million characters on those plans at Speechify's standing price.

### 5. Typecast: Best for a clone from as little as 5 seconds through an API

**What does it do?** Typecast's API makes an instant clone from one WAV or MP3 of 5 to 150 seconds, up to 25 MiB, and returns a usable voice. Instant cloning reached the API in June 2026 and professional cloning, from 5 minutes to 3 hours of audio, in August 2026. API Lite at $15 a month includes 50 custom voice slots shared between instant and professional clones, and the Basic app plan at $5 a month includes one instant cloning slot.

**Who is it for?** Typecast suits a developer who wants cloning from a short sample on a low monthly plan and will accept a contract as the consent step.

<div class="cmp-proscons" data-rival="typecast"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Instant clones from 5 seconds of audio</li><li>A $5 app plan and a $15 API plan that both include cloning</li><li>Professional cloning through the same API since August 2026</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Consent is a contractual warranty, and Typecast may ask for evidence of it</li><li>The API terms let Typecast train and improve its models on submitted voice data</li><li>Cloned voices can't be exported</li></ul></div></div>

**Pricing:** API Lite at $15 a month for 200,000 credits at $0.075 per 1,000 credits, with overage at $0.09 per 1,000, and API Plus at $280 a month for 4 million credits.

### 6. Gradium: Best for a 10-second clone on a real-time voice agent API

**What does it do?** Gradium sells real-time speech for voice agents and clones a voice from 10 seconds of clear audio, with new clones enhanced by default since 23 October 2026. Gradium's Free plan includes 6 custom voices and its $12 XS plan 1,000, with commercial use from XS. Gradium says text to speech starts in under 50ms and returns 8 kHz mu-law for phone calls.

**Who is it for?** Gradium suits a team building a voice agent in English, French, German, Spanish, or Portuguese that wants a clone on the same API as its speech.

<div class="cmp-proscons" data-rival="gradium"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>A 10-second clone, with 6 custom voices on a $0 plan</li><li>A WebSocket that takes text token by token</li><li>Mu-law and A-law output at 8 kHz</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Five languages</li><li>A professional clone starts on the $312 M plan, with a one-time training fee of 1 million credits</li><li>Commercial use isn't on the Free plan</li></ul></div></div>

**Pricing:** 1 credit per character, with plans from Free to L at $1,480 a month and rates from $58 per million credits on XS to $36 on L.

## What does a 10-second clone cost to make and keep?

Making a clone costs from nothing to $15 a month, and keeping one costs differently at each company. Gradium's Free plan, Bland Speech's pay as you go billing, and Inworld's $0 On-Demand plan start cloning without a monthly fee, while ElevenLabs starts at $6 a month, Speechify at $10, and Typecast at $5 for the app and $15 for the API. MiniMax charges $1.5 once per voice and removes a voice that sits unused for a week.

At $40 per million characters, Bland Speech sits between Speechify's $10 to $15 and MiniMax's $60 to $100, and below ElevenLabs' $80 models. Bland's [voice cloning API pricing guide](/blog/voice-cloning-api-pricing) compares the full bills.

## How were these voice cloning tools compared?

Samuel Oram-Jones compared six tools on the audio an instant clone needs, the plan or charge that gates it, the sample rules, the consent step, and the price of speech. Every rival detail comes from that company's own pricing, documentation, and terms pages.

Bland's [text to speech comparison hub](/compare/text-to-speech) covers the same companies by job, and the [voice cloning comparisons](/compare/text-to-speech/voice-cloning) go one provider at a time. A team deciding how much recording to collect can read [how much audio it takes to clone a voice](/blog/how-much-audio-to-clone-a-voice), and a team weighing the clone types can read [instant vs professional voice cloning](/blog/instant-vs-professional-voice-cloning).

## Frequently asked questions

### What is the best voice cloning from 10 seconds of audio?

Bland Speech is the best for a clone that will speak on phone calls. One sample of about 10 seconds makes an instant clone, and Bland Speech v3 ranked first of 13 AI voices on Audio Realism Bench on 4 August 2026. Inworld, Typecast, and Smallest AI ask for less audio.

### Can a voice be cloned from 10 seconds of audio?

Yes. Bland Speech, ElevenLabs on Eleven v4, MiniMax, Speechify, and Gradium all document 10 seconds as enough for an instant clone, and Cartesia says 10 seconds is enough to get started.

### Does a longer sample make a better clone?

At several providers it does. Cartesia says up to 60 seconds better retains accent, Inworld says similarity improves up to its 30-second cap, and Fish Audio says a minute or two improves fidelity. Bland says the quality of the sample sets the ceiling, and asks for one clean take of about 10 seconds.

### What is the shortest sample a voice cloning API accepts?

Inworld clones from as little as 3 seconds and Typecast and Alibaba Cloud's Qwen voice enrollment from 5 seconds. NVIDIA's Magpie TTS Zeroshot takes 3 to 10 seconds but needs access approval.

### Is a 10-second clone good enough for a business?

A 10-second clone is enough to start, and a professional clone is the next step when a business wants more fidelity. Bland Speech trains a professional clone on 30 minutes or more of audio after a consent and rights review, as ElevenLabs does from about 30 minutes.
