Description
Real-time voice AI mostly lives in the cloud, billed by the minute with everything you say uploaded. Parlor is fully on-device, real-time multimodal AI similar to GPT-Live, listening, seeing and answering at once.
It runs Gemma via MLX and LiteRT-LM on Apple Silicon with Kokoro for speech, and the author uses it to offer free English speaking practice to hundreds of monthly users from a home server.
Fully local:Speech in, understanding and speech out on device.
Multimodal:Voice and vision together.
Real time:Low-latency conversation.
Apple Silicon:Optimized with MLX.
It runs Gemma via MLX and LiteRT-LM on Apple Silicon with Kokoro for speech, and the author uses it to offer free English speaking practice to hundreds of monthly users from a home server.
Features
Fully local:Speech in, understanding and speech out on device.
Multimodal:Voice and vision together.
Real time:Low-latency conversation.
Apple Silicon:Optimized with MLX.
