LocalAI

LocalAI Realtime API is a self-hosted, OpenAI-compatible realtime interface for low-latency voice and text conversations. It is built for developers who want to serve multimodal chat locally or on their own infrastructure rather than relying on a hosted API.

To get started, you define a pipeline model in a YAML configuration file and wire together the components for voice activity detection, transcription, language model inference, and text-to-speech. The docs show both WebSocket and WebRTC transports, so you can use it for backend integrations or browser-based voice apps.

This is infrastructure, not a turnkey assistant. The realtime experience depends on the models and backends you install and configure, so quality and latency will vary based on your stack. WebRTC also requires the Opus backend to be installed separately.

LocalAI also documents authentication and authorization options, including API keys, OAuth/OIDC, role-based access, and per-user usage tracking, which makes it more suitable for multi-user deployments than a simple local demo. Pricing was not publicly specified in the content provided, and the exact setup requirements beyond model configuration were only partially documented here.