Tavus launches Phoenix-4.5, claiming the most expressive real-time human rendering model yet
LinusEkenstam · x · 2026-09-11
Tavus introduced Phoenix-4.5, billed as the fastest and most expressive real-time human rendering model on the market. The update brings richer facial animation, more expressive emotion and micro-movements, and rendering that now extends through the upper body. The company claims it's the closest AI has come to passing a face-to-face Turing test.
More from Multimodal
- GPT Astra + Blender previs, then Seedance 2.5: a precise AI video workflow — Mr_AllenT · 2026-09-11
- Kirby meets The Truman Show: fan video made with MiniMax H3 across 30 workflow files — dramaton42 · 2026-09-11
- ComfyUI VDN-H3 24GB v1.1.0 fixes adapter mapping, boosts prompt following at ~4% speed cost — NoMouse9610 · 2026-09-11
- Seedance 2.0/2.5 vs Kling 3.0 Tested on Character and Camera Consistency — letandrewcook · 2026-09-11
- Dreamina-Generated Haute Couture Video Shows Off AI Fashion Visuals — flexeble2020 · 2026-09-11
- Lock a realistic base frame first: prompt template for photoreal GPT Image 2.5 output — techhalla · 2026-09-11