MarkTechPost Tests Seven Voice Cloning APIs on Similarity, Consent and Price

The 10-second reference was one speaker in a quiet room, saved as WAV, and every platform read the same two sentences, one conversational and one containing numbers and a proper name. Each vendor's instant cloning path was used with default settings. Two vendors were handled differently. Hume documents 15 seconds as its floor, so its clone used the same speaker extended to 15 seconds, and ElevenLabs recommends 1 to 2 minutes for Instant Voice Cloning, so its 10-second clone ran below vendor guidance.

Hume published its Voice Replication Leaderboard on September 10, 2026. It tested 11 models on 25 reference voices with 7 prompts. Three blind raters scored each clip from 1 to 5 on sounding like the reference. Among the vendors in the comparison, Fish Audio s2-pro led at 4.03. Cartesia sonic-3.5 scored 3.70, ElevenLabs Multilingual v2 scored 3.68, Cartesia sonic-3.6-beta scored 3.63, and Inworld TTS-2 scored 3.62. ElevenLabs Eleven v3 ranked last of 11 at 2.91. The board also showed why one number can mislead: Cartesia sonic-3.6-beta topped naturalness at 4.36 yet ranked eighth on identity, while Inworld TTS-2 topped audio quality at 4.61. Full per-category data sits on the Hugging Face leaderboard.

ElevenLabs offers Instant Voice Cloning and Professional Voice Cloning. Its documentation recommends 1 to 2 minutes of clean audio for instant cloning, while professional cloning needs at least 30 minutes, ideally 2 to 3 hours, on the Creator plan or above. It also has the strictest gate in the comparison: the voice owner reads on-screen text aloud through Voice Captcha. The listed commercial license starts with Starter at $6 per month, and language coverage is more than 70 for Eleven v3. List prices range from $50 for Flash to $100 for v3 per one million characters. API rates are $0.10 per 1,000 characters on Eleven v3 and Multilingual v2, and $0.05 for Flash, Turbo and v3 Conversational. Consent involves rights attestation for instant cloning and Voice Captcha for professional cloning.

Cartesia clones from 10 seconds of audio. Sonic 3.6 and newer use up to 60 seconds to hold accents better. Pro Voice Clones train on 30 or more minutes and start on the $49 Startup plan. Sonic bills 1 credit per character, and plans run from $5 for 100,000 credits to $299 for 8 million credits. Cartesia lists 44 languages. Its consent requirement is permission required by terms, its commercial license starts with Pro at $5 per month, and its list price is about $37 to $50 per one million characters.

Inworld instant clones work from 3 seconds, and samples up to 30 seconds improve similarity. Cloning itself is free. Realtime TTS-2 lists $25 per one million characters on demand, falling to $15 at $300 a month and $12.50 at $1,500. TTS-2 Flash starts at $15 and falls to $7. Professional cloning is in beta, needs 10 minutes of audio, and supports English only. Inworld supports more than 200 languages and locales, with 15 in its top quality tier. Its consent requirement is rights confirmation, its commercial license comes from the free On-Demand tier, and its listed price is $12.50 to $25 for TTS-2.

Gradium, founded by Kyutai co-founders, clones from 10 seconds. The free tier includes 5 clones for non-commercial use. Paid plans start at $13 for 225,000 credits, billed at 1 credit per character. Pro Voice Clone needs 30 minutes minimum and starts on the $340 M plan. Coverage is narrow: English, French, German, Spanish and Portuguese. Its listed reference is 10 seconds for instant cloning and 30 minutes for professional cloning, with 2 hours recommended, and its list price is about $36 to $58 per one million characters. Owner consent is required by policy, and the commercial license starts with XS at $13 per month.

Fish Audio bills its API at $15 per one million UTF-8 bytes. English text runs about 1 byte per character, while Chinese, Japanese and Korean run about 3, which roughly triples the effective price. S2.1 Pro covers 83 languages. Instant clones work from about 10 seconds. Professional clones take 10 to 180 minutes and require live ownership verification. Commercial use starts on the $15 Plus plan and covers verified voices the user owns. Its consent method is a live ownership check for pro clones.

Resemble AI Rapid Clone needs 10 seconds. Its reference range is listed as 10 seconds for instant cloning and 10 to 25 or more minutes for professional cloning. Professional Clone uses verifiable consent, and the cloning API requires the Business plan. Resemble AI lists 23 languages. Its price page lists no TTS rate; trackers reported $0.0005 per second in mid-2026, converted at about 1,000 characters per minute, for an estimate of about $30 per one million characters.

Hume documents 15 seconds as its reference floor, with no professional reference documented. It asks users to upload from a consenting speaker. Commercial licensing starts with Creator at $14 per month, and it lists 11 languages. Its list price is $50 to $150 by plan, and credit-plan rates equal the plan price divided by included characters.