Artificial Analysis Launches Controlled Voice Arena TTS Leaderboard
Artificial Analysis has launched the Controlled Voice Arena, a text-to-speech (TTS) leaderboard featuring a public voting arena. To isolate model capability from voice preference, the evaluation uses 8 uniformly cloned voices to test different TTS models. The organization has also published its methodology and the cloned voice samples used for testing.
Accent and Open-Weight Rankings
In the latest results, Cartesia Sonic 3.5 took the first place in both British and American accent evaluations. For the British accent, it was followed by Inworld Realtime TTS-2, ElevenLabs Eleven v3, and MiniMax Speech. For the American accent, ElevenLabs Eleven v3 and Inworld Realtime TTS-2 ranked closely behind, with Soniox TTS Real-Time v1 and MiniMax also making the top list. In the open-weight category, Fish Audio S2 Pro currently holds the top spot, followed by Mistral Voxtral TTS, Resemble AI Chatterbox, Microsoft VibeVoice 7B, and OpenVoice v2.
2026-07-09 ~ 2026-07-09 · 5 related posts
- [source] Unified Voice Timbre Leaderboard Released — ArtificialAnlys · 2026-07-09
- American Accent TTS Leaderboard Updated — ArtificialAnlys · 2026-07-09
- [source] British Accent TTS Leaderboard Updated — ArtificialAnlys · 2026-07-09
- [source] Open-Source Voice Model Rankings Updated — ArtificialAnlys · 2026-07-09
- Artificial Analysis Launches Controlled Voice Synthesis Leaderboard — ArtificialAnlys · 2026-07-09