Drama 3 Preview leads TTS pronunciation robustness: two firsts, one second across four categories
ArtificialAnlys · x · 2026-10-14
Artificial Analysis details its Pronunciation Robustness benchmark, which measures whether TTS models correctly pronounce challenging text across four categories, judged by human reviewers against pre-agreed pronunciations. Drama 3 Preview leads overall at 93.2%.
- Standalone Terms: #1 at 95.9%, ahead of Eleven v3 Conversational (95.1%) and Gemini 3.1 Flash TTS (93.7%)
- Preserving Exact Sequences: #1 at 86.3%, ahead of SpaceXAI TTS (85.7%) and HeyGen Voice (83.4%)
- Expanding Shorthand: #2 at 93.6%, behind Eleven v4 (94.1%), ahead of Eleven v4 Turbo (91.9%)
- Contextually Appropriate: 95.3%, with Gemini 3.8 Flash TTS leading the category at 97.9%
Related event: Fish Audio Open-Sources Drama 3, Tops Open-Weight TTS Leaderboard(12 posts)→
More from Multimodal
- AI video: Tony Soprano convinces Anakin to turn to the dark side — indiegameplus · 2026-10-14
- Designer: new model is highly capable at design but defaults to generic 'AI slop' look unsteered — zacharynado · 2026-10-14
- Krea 3 lands Thursday: 2K images in 7s, 3.7x faster than Nano Banana Pro — TitusTeatus · 2026-10-14
- Gemini 4 Argon turns your webcam feed into a playable synthesizer app — GeminiApp · 2026-10-14
- Strikingly lifelike AI short film imagines Hannibal before he was famous — aeeiee · 2026-10-14
- Prime Intellect acquires Moondream to build efficient multimodal models — Sentdex · 2026-10-14