Anam blind benchmark: 68% positive lip-sync ratings vs ~40% for rivals on 960 conversations
SucceededMind · x · 2026-10-21
Anam launched Cara-4.1 Express, a real-time interactive AI avatar model, with a blind benchmark: across 60 raters and 960 conversations, it scored 68% positive lip-sync ratings versus roughly 40% for the next best competitor, ranking highest across all six categories with better eye contact and micro-expressions. A new Director Notes feature guides avatar performance.
Commenters note that badly timed mouth movements remain the fastest way to break an AI agent's illusion, and this gap looks real rather than marketing fluff. Anam claims adding a face to agents boosts engagement by 44%, offering APIs for support, sales, tutoring, and medical use cases.
Related event: Anam Unveils Avatar Cara-4.1 with 1.2s Latency, Beats Rivals in Blind Tests(2 posts)→
More from Multimodal
- $33.49 Claude-made hype video remake landed a $450k deal — full open-source pipeline — zealcaiden · 2026-10-21
- "The Weight of Silence": AI-generated black-and-white series visualizes unspoken pain — LudovicCreator · 2026-10-21
- AssemblyAI to Preview TTS-1, a Text-to-Speech Model Built for Voice Agents — AssemblyAI · 2026-10-21
- AI-generated short imagines 2126: billionaires atop kilometer-high towers, the poor below — KaptanHindustan · 2026-10-21
- Midjourney sref Code 1080457169 Nails Pointillist Snow-Scene Oil Painting Style — sergeantsref · 2026-10-21
- Moss-Image SFT + New DMD2 LoRA: Sharper 8-Step Generation, But Style Collapse — AgeNo5351 · 2026-10-21