Parlor drops local real-time multimodal AI

Parlor demonstrates that "Her-like" multimodal interaction is viable on consumer hardware today. By utilizing Gemma 4 E2B for simultaneous reasoning and local inference for privacy, it eliminates per-token costs. The combination of barge-in support, sentence-level TTS streaming, and built-in VAD removes the friction of voice interfaces, making it ideal for low-latency applications like language learning.