A recent evaluation of voice cloning APIs compared 11 models based on speaker similarity, consent verification, and pricing. The study utilized a 10-second reference audio clip for testing across various platforms, with some vendors requiring longer samples for optimal performance.
The Hume Voice Replication Leaderboard, published on September 10, 2026, ranked Fish Audio s2-pro as the top model for speaker similarity with a score of 4.03. Other notable performers included Cartesia sonic-3.5 at 3.70 and ElevenLabs Multilingual v2 at 3.68.
Pricing structures vary significantly among providers. ElevenLabs charges $0.10 per 1K characters for its primary models, while Inworld offers TTS-2 at $25 per 1M characters. Fish Audio bills $15 per 1M UTF-8 bytes, and Cartesia offers plans ranging from $5 to $299.
Professional cloning requirements differ by vendor, often necessitating 10 to 30 minutes of audio and explicit consent verification. For instance, ElevenLabs requires a Voice Captcha for professional cloning, while Resemble AI emphasizes deepfake detection and requires a Business plan for its API.
