AssemblyAI launches its first TTS model, entering top 12 in both voice arenas
ArtificialAnlys · x · 2026-10-29
AssemblyAI has launched its first Text-to-Speech model, AssemblyAI-TTS-1, which enters the top 12 of Artificial Analysis's Provider Voice Arena and the top 10 of Controlled Voice Arena.
- Provider Voice Arena: Elo 1227, #12 overall (Customer Service #13, Knowledge Sharing #13, Assistants #13, Entertainment #17); #12 in both US and UK English.
- Speed/price: 162 chars/sec and $15/1M characters — among the fastest at a mid-tier price.
- Pronunciation robustness: 81.7% overall (#16), with shorthand expansion (73.8%) as the weak spot.
Positioning: competitive on speed and price, upper-mid quality but not top-tier.
Related event: AssemblyAI Launches Its First TTS Model, TTS-1(5 posts)→
More from Multimodal
- Retro Anime Style LoRA Recreates 80s/90s Cel-Shaded Look with Trigger Word and Weight Tips — vizsumit · 2026-10-29
- AI Short Film 'Aliens Picked the Wrong House' Turns a Humble Pie Into Comedy — 47studios · 2026-10-29
- Solo Creator Makes Original Anime 'Dragonshard' with AI, Shows Sorcerer Transformation — AshfallAnimation · 2026-10-29
- AI short film "Hunt or Be Hunted" created with Seedance 2.5 — Kyrannio · 2026-10-29
- Creator Uses AI to Shoot Horror Short '3:17 — She Knew His Name' Set in His Florida Neighborhood — Ghostatlasstories · 2026-10-29
- Runway Dev launches on Codex for autonomous API integration — runwayml · 2026-10-29