- Published on
Best Voice Cloning APIs in 2026: Speaker Similarity, Consent Checks, and Price per 1M Characters
- Authors
- Name
- DeepHorus
Voice cloning providers in 2026 face a harder test than good audio: sounding like one specific person. A new comparison of seven platforms cloned a single 10-second reference clip on every service. Results show wide gaps in similarity, consent checks, and price.
Testers recorded one speaker in a quiet room and used each vendor's instant cloning path. Hume required 15 seconds; ElevenLabs recommends 1 to 2 minutes, so those clones ran outside the standard. Prices were checked on September 20, 2026.
Hume's Voice Replication Leaderboard, published September 10, 2026, tested 11 models on 25 reference voices with 7 prompts. Three blind raters scored each clip from 1 to 5. Fish Audio's s2-pro led at 4.03, Cartesia sonic-3.5 scored 3.70, and ElevenLabs Multilingual v2 scored 3.68. Eleven v3 ranked last at 2.91.
The leaderboard also shows why one number misleads. Cartesia sonic-3.6-beta topped naturalness at 4.36 yet ranked 8th on identity, while Inworld TTS-2 topped audio quality at 4.61. Buyers need category-level data, not just an average.
Pricing varies sharply. ElevenLabs charges $0.10 per 1K characters on Eleven v3 and Multilingual v2, with Flash and Turbo at $0.05. Inworld lists $25 per 1M characters on demand, falling to $12.50 at higher tiers. Fish Audio bills $15 per 1M UTF-8 bytes, tripling effective price for Chinese, Japanese, and Korean.
Consent and commercial terms differ. ElevenLabs requires the voice owner to read on-screen text through Voice Captcha. Fish Audio demands live ownership verification for professional clones; commercial use starts on its $15 Plus plan. Gradium's free tier allows five clones for non-commercial use only.
Minimum audio ranges from Inworld's 3 seconds to ElevenLabs' 1-2 minutes. Teams must weigh similarity, naturalness, price, and language coverage. The side-by-side test and public leaderboard give developers a clearer basis for that decision.
