Two well known voice cloning platforms disappeared in under six months. PlayHT shut down on December 31, 2025, after Meta acquired its parent company PlayAI, and every stored voice clone was deleted. LOVO AI filed for Chapter 7 bankruptcy on May 27, 2026. Anyone who built a podcast, course or product on those voices had to start again.
The tools that replaced them clone a voice from far less audio. Cartesia needs 3 seconds, Resemble AI about 5 seconds and Fish Audio 10 seconds, while API prices for an hour of synthetic speech now range from under $1 to about $5.40. The rules changed too: EU AI Act deepfake disclosure applies from August 2, 2026, and the US NO FAKES Act cleared the Senate Judiciary Committee in June 2026.
This list covers the 12 voice cloning tools worth using in 2026 and five alternatives to cloning, with the sample length each tool needs, its consent safeguards, real prices and our own cost per hour comparison.
Quick picks:
- Best voice quality: ElevenLabs
- Cheapest API per hour: Fish Audio
- Fastest real time voice agents: Cartesia
- Emotional delivery: Hume AI
- Free and self hosted: Chatterbox
- Podcast editing with your own voice: Descript
- Video dubbing in your voice: Rask AI
- Best alternative to cloning: AI voice design
Tools Compared Side by Side
Voice Cloning Sample Length, Entry Price and Safeguards (2026)
| Tool | Audio Needed to Clone | Entry Price With Cloning | Consent or Safety Feature |
|---|---|---|---|
| ElevenLabs | About 1 minute (instant), 30 minutes (professional) | $6 a month | Voice CAPTCHA for professional clones |
| Fish Audio | 10 to 30 seconds | Free plan | Basic terms |
| Cartesia | 3 seconds | $5 a month | Enterprise controls |
| Hume AI | Short sample | $14 a month for full use | Commercial license on paid plans |
| MiniMax Audio | Short sample | $1.50 per cloned voice | Pay per voice |
| Resemble AI | About 5 seconds | $2 a month per voice | Watermarking and deepfake detection |
| Chatterbox | A few seconds | Free, self hosted | Built in audio watermark |
| Descript | About 60 seconds | $16 a month billed yearly | Recorded consent statement |
| Speechify Studio | About 10 seconds | $19 a month | Commercial cloning on Premium+ |
| Murf AI | About 2 minutes | Enterprise only | SOC 2 Type II, ISO 27001, HIPAA |
| Rask AI | Pulled from your uploaded video | $28 a month billed yearly | Clones stay inside Rask |
| Kukarella | Short sample | $21 a month, one clone | Basic terms |
The 12 Latest AI Voice Cloning Tools
1. ElevenLabs
| Website | elevenlabs.io |
| Best for | Audiobooks, YouTube narration and multilingual content |
| Free plan | 10,000 credits a month, cloning excluded |
| Paid plans | Starter $6, Creator $22, Pro $99, Scale $299, Business $990 a month |
| Cloning | Instant from Starter, professional from Creator with voice verification |
ElevenLabs launched its v4 and v4 Turbo models on September 28, 2026, covering 90+ languages. A professional clone needs about 30 minutes of clean audio and a spoken voice CAPTCHA, and the platform only lets you build one from your own voice, per Fora Soft. Included minutes cost about $0.165 to $0.20 each on subscriptions, and API rates run $0.04 to $0.08 per 1,000 characters.
2. Fish Audio
| Website | fish.audio |
| Best for | High volume narration on a tight budget |
| Free plan | 8,000 credits a month with instant cloning |
| Paid plans | Plus $15 a month or $11 billed yearly; API at $15 per million characters |
| Cloning | 10 to 30 seconds of audio, 80+ languages |
Fish Audio has the lowest API price on this list, at about $0.81 per hour of speech. It previewed its Drama 3 model in September 2026 and supports 60+ emotion tags with streaming latency under 300 milliseconds. Check its commercial terms closely, since its free plan rules differ between pages.
3. Cartesia
| Website | cartesia.ai |
| Best for | Real time voice agents and phone bots |
| Free plan | 20,000 credits a month |
| Paid plans | Pro $5, Startup $49, Scale $299 a month; voice agent calls at $0.06 a minute |
| Cloning | Instant from 3 seconds on Pro, professional from Startup |
Cartesia built its Sonic model for speed, with about 90 milliseconds to the first audio, per Dupple. That latency suits customer service agents and AI receptionists, where a delay above half a second breaks the conversation. Its 3 second sample bar is the lowest on this list.
4. Hume AI
| Website | hume.ai |
| Best for | Emotionally expressive reads and empathic voice agents |
| Free plan | 10,000 characters, clone creation for testing only |
| Paid plans | Starter $3, Creator $14, Pro $70 a month |
| Cloning | Unlimited cloned voices from Creator, API access on Pro |
Hume AI trains its Octave models to read emotion from text, so a clone sounds worried, excited or calm based on the script. Its Empathic Voice Interface adds 40 to 1,200 minutes of conversational agent time depending on plan. Pro overage costs $0.05 per 1,000 characters, about $2.70 an hour.
5. MiniMax Audio
| Website | minimax.io |
| Best for | Developers paying per voice and per character |
| Free plan | Trial credits |
| Paid plans | Speech 2.8 HD at $100 per million characters, Turbo at $60 |
| Cloning | $1.50 per rapid cloned voice, $3.00 per Voice Design creation |
MiniMax charges per cloned voice, which suits apps that create hundreds of custom voices for users. Its Voice Design feature builds a new synthetic voice from a text description, a safe option for brands that want a unique voice with zero real person attached. Pricing comes from the August 2026 comparison by InVideo.
6. Resemble AI
| Website | resemble.ai |
| Best for | Enterprises that need cloning plus deepfake detection |
| Free plan | Flex pay as you go at $0 base |
| Paid plans | About $0.0005 per second of speech; rapid clones $2 and professional clones $5 a month per voice |
| Cloning | About 5 seconds for a rapid clone, 20+ languages |
Resemble AI now leads with deepfake detection, and its Detect product flags synthetic audio in call centers and newsrooms. It offers SOC 2 Type 2 compliance, SSO and on premise hosting. At $0.0005 a second, an hour of speech costs about $1.80.
7. Chatterbox
| Website | github.com/resemble-ai/chatterbox |
| Best for | Developers who want free, private, self hosted cloning |
| Free plan | Free under the MIT license |
| Paid plans | Your own GPU and compute costs |
| Cloning | Zero shot from a few seconds, 23+ languages in the multilingual model |
Resemble AI released Chatterbox as open source, and every output carries an imperceptible watermark that detection tools read. Emotion intensity control lets one cloned voice move from flat narration to dramatic delivery. Resemble reports its hosted Chatterbox beat ElevenLabs in 63.75% of blind tests, a vendor claim worth testing on your own scripts.
8. Descript
| Website | descript.com |
| Best for | Podcasters and video editors fixing words in their own recordings |
| Free plan | 60 media minutes a month with limited cloning and a watermark |
| Paid plans | Hobbyist $16, Creator $24, Business $50 a month billed yearly |
| Cloning | About 60 seconds plus a recorded consent statement, English only |
Descript lets you type a correction into a transcript and hear it in your own voice, so a misspoken product name gets fixed in seconds. It added an ElevenLabs voice library on September 17, 2026. Free and Creator plans cap cloned speech to a 1,000 word vocabulary.
9. Speechify Studio
| Website | speechify.com |
| Best for | Creators who already use Speechify for reading |
| Free plan | 600 Studio credits, cloning excluded |
| Paid plans | Studio Starter $19 a month or $100 a year; Studio $19 to $49 per user a month |
| Cloning | About 10 seconds, commercial use on Premium+ |
Speechify Studio runs as a separate subscription from the Speechify reading app, so check which plan you pay for before cloning. Its Simba 3.2 model shipped in July 2026. Mobile first creators get a quick path from phone recording to cloned voiceover.
10. Murf AI
| Website | murf.ai |
| Best for | Corporate training and e-learning teams |
| Free plan | Short trial |
| Paid plans | Creator $19 a month billed yearly, Business $66, Enterprise custom |
| Cloning | Enterprise only, built with the Murf team from about 2 minutes of audio |
Murf gates cloning behind Enterprise contracts, which gives regulated companies a reviewed and documented process. It carries SOC 2 Type II, ISO 27001, HIPAA and GDPR compliance, and its Falcon API starts audio in under 100 milliseconds. Learning teams use it to keep one consistent trainer voice across hundreds of modules.
11. Rask AI
| Website | rask.ai |
| Best for | Dubbing videos into other languages in the original speaker voice |
| Free plan | 3 minute trial |
| Paid plans | Creator $39 a month or $28 billed yearly, priced per minute of media |
| Cloning | Pulled from the uploaded video, 32 languages for cloned voices |
Rask AI clones the speaker straight from the source video, then translates and dubs it into 135+ languages, 32 of them with the cloned voice. A YouTube creator turns one English video into Spanish, Hindi and Portuguese versions in their own voice. Cloned voices stay inside Rask and export only as finished audio.
12. Kukarella
| Website | kukarella.com |
| Best for | Multilingual voiceovers across rare languages |
| Free plan | Signup credits, 10 requests, personal use only |
| Paid plans | Prime $21 a month with one cloned voice |
| Cloning | Short sample, 152 languages |
Kukarella covers 152 languages, the widest range on this list, which helps NGOs and educators producing content in regional languages. It bundles one cloned voice into the Prime plan. Pricing details come from the October 2026 roundup by Notevibes.
Alternatives to AI Voice Cloning
Cloning fits when you own the voice or hold written consent. When you need a voice that belongs to zero real person, or a human performance, these five alternatives cover the same jobs.
Voice Cloning Alternatives Compared (2026)
| Alternative | How It Works | Typical Cost | Best For |
|---|---|---|---|
| AI voice design | Describe a voice in text and the model builds a new synthetic one | $3 per voice on MiniMax; included in ElevenLabs plans | Brand voices tied to zero real person |
| Stock AI voice libraries | Pick a ready made voice from thousands of presets | Included in TTS plans from $5 to $22 a month | Fast narration and prototypes |
| Licensed voice marketplaces | Use a real voice the owner shares for a fee | Custom pricing on Respeecher; royalty based on ElevenLabs | A real human sound with legal clearance |
| Speech to speech conversion | Record the delivery yourself and convert it to another voice | Custom pricing for real time call center use | Acting, timing and emotion control |
| Human voice actors | Hire a professional to record the script | $200 to $400 per finished hour for audiobooks; $350 to $750 for corporate narration | Ads, premium audiobooks and union projects |
1. AI Voice Design
Voice design builds a brand new voice from a text prompt such as a calm female narrator with a light Irish accent. ElevenLabs Voice Design works with its v3 and v4 models and supports audio tags, per the ElevenLabs help center, and MiniMax charges $3.00 per Voice Design creation. A designed voice belongs to zero living person, which removes consent and right of publicity risk entirely.
2. Stock AI Voice Libraries
Every tool on this list ships preset voices that work from day one with zero sample audio. Cartesia, ElevenLabs, Fish Audio and Murf each offer hundreds to thousands of ready voices across ages, accents and languages. A preset voice suits a product demo or internal training video where a unique brand voice adds little value.
3. Licensed Voice Marketplaces
Marketplaces let voice owners license their own clones to others. Respeecher runs a Voice Marketplace with 40+ AI voices on custom pricing, per Toolradar, and the ElevenLabs Voice Library pays voice owners when others use their shared professional clones. Each voice arrives with permission already in place.
4. Speech to Speech Conversion
Speech to speech keeps your timing, pauses and emotion, then swaps the timbre for another voice. Respeecher sells real time conversion for call centers, and film and game studios use the same approach to match an actor across languages or ages. Pick it when delivery matters more than speed of production.
5. Human Voice Actors
A professional narrator charges $200 to $400 per finished hour for audiobooks and $350 to $750 per finished hour for corporate narration, per 2026 rate data compiled by VoxBooster. ACX style royalty deals cost $0 upfront in exchange for 20% of net sales. Actors also license digital replicas of their voice, with capture sessions starting around $1,500.
Price gap in practice: a 10 hour audiobook costs $2,000 to $4,000 with a professional narrator, about $42 on ElevenLabs Multilingual at API rates and about $8 on Fish Audio. Premium titles and union projects still go to human narrators, while backlist and course content moves to AI.
What One Hour of Cloned Voice Costs: Our Analysis
Subscription prices hide the true cost of audio. We converted each published 2026 API rate into one hour of speech, using 150 words a minute and about 54,000 characters per hour.
API Cost per Hour of Synthetic Speech (2026 List Rates)
| Model or Plan | Published Rate | Cost per Hour |
|---|---|---|
| Fish Audio API | $15 per million characters | About $0.81 |
| Resemble AI Flex | $0.0005 per second | About $1.80 |
| ElevenLabs Flash or Turbo API | $0.04 per 1,000 characters | About $2.16 |
| Hume AI Pro overage | $0.05 per 1,000 characters | About $2.70 |
| MiniMax Speech 2.8 Turbo | $60 per million characters | About $3.25 |
| Cartesia voice agent calls | $0.06 per minute | About $3.60 |
| ElevenLabs Multilingual API | $0.08 per 1,000 characters | About $4.20 |
| MiniMax Speech 2.8 HD | $100 per million characters | About $5.40 |
A 10 hour audiobook costs about $8 on Fish Audio and about $42 on ElevenLabs Multilingual at API rates. Quality, language coverage and voice likeness decide the right pick, so test the same script on two or three tools before committing a full project.
Discontinued Platforms to Avoid
Voice Cloning Platforms That Shut Down
| Platform | What Happened | Date |
|---|---|---|
| PlayHT | Parent company PlayAI acquired by Meta; API offline July 26, 2025; all clones deleted at shutdown | December 31, 2025 |
| LOVO AI (Genny) | Filed Chapter 7 bankruptcy | May 27, 2026 |
Keep your original training recordings and export generated audio regularly, since a shutdown deletes stored clones, per Cognitive Future. Watch out for playhtai.com, an unaffiliated site using the PlayHT name.
Voice Cloning Laws and Fraud Risks in 2026
- EU AI Act Article 50 requires deepfake disclosure from August 2, 2026, and providers of existing generative systems must add machine readable marking by December 2, 2026.
- The NO FAKES Act of 2026 (S.4591) cleared the Senate Judiciary Committee unanimously in June 2026 and awaits a floor vote. It proposes statutory damages of $5,000 to $25,000 per unauthorized replica.
- Tennessee ELVIS Act protects musicians and individuals against unauthorized voice clones under state law.
- The FCC ruled in February 2024 that AI generated voices in robocalls fall under the Telephone Consumer Protection Act, making them illegal unless the person called gave prior consent.
- US consumers lost a record $15.9 billion to fraud in 2025, per FTC testimony in March 2026, and Pindrop recorded a 680% rise in deepfake voice activity in its 2025 report.
Clone only your own voice or a voice you hold written permission to use, keep that consent on file and label synthetic audio in published content.
Frequently Asked Questions
Which Tool Needs the Least Audio?
Cartesia, which clones a voice from 3 seconds of audio. Resemble AI needs about 5 seconds and Fish Audio 10 seconds.
What Is the Best Free Option?
Chatterbox for self hosting under the MIT license, and Fish Audio for a hosted free plan with instant cloning.
What Replaced PlayHT?
Most former PlayHT users moved to ElevenLabs for quality, Fish Audio for price or Cartesia for real time voice agents.
Is AI Voice Cloning Legal?
Cloning your own voice or a voice you hold consent for is legal in most places. An unauthorized clone of another person breaks right of publicity laws such as the Tennessee ELVIS Act, and AI voices in robocalls are illegal under the FCC 2024 ruling.
Find more YouTube tools:
