// blog
Parley: two personas, one conversation
Pair up two of your channels — bridges, direct LLM connections, or one of each — give them personas, and watch them talk. Primeta now stages two-way AI conversations with synced avatars, voices, and a shared transcript.
Chat with your personas directly — no MCP client needed
Primeta now supports direct-LLM chat using your own OpenAI, Anthropic, xAI, or Ollama Cloud key. Paste a key, pick a model, chat. No Claude Code, no local agent, no MCP bridge required.
Release notes: voice, eyes, breath, and everything we shipped this week
A week of avatar improvements — lipsync mouth that actually moves, eyes that dart around, breathing that you can see, emotion tags that match face and voice, and a pile of small TTS latency wins under the hood.
Animation File Formats for 3D Avatars: FBX, VRMA, BVH, glTF, and .anim Compared
A practical guide to the animation file formats used in 3D avatar systems — FBX, VRMA, BVH, glTF, and Unity .anim — how they work, when to use each, and how retargeting maps motion data onto VRM models.
From Phonemes to Mouth Shapes: Animating VRM Models
A technical look at how text-to-speech phoneme timelines drive VRM blend shape animation in real time — from the 5 core visemes to exponential interpolation, emotion conflicts, and FFT frequency fallback.
What Are Phonemes?
Phonemes are the smallest units of sound in language that distinguish meaning — and they're crucial for making AI avatars speak naturally. Understanding how phonemes work helps explain why lip sync technology requires more sophistication than simply matching text to mouth movements.