One of the most praised features in our private alpha builds is our completely silent, bot-free meeting transcription engine. Here is a technical peek into how our audio driver integration works under the hood.
By capturing audio directly from your OS kernel sound buffer, Maple transcribes calls with zero participant alerts and zero network latency.
When you take a call on Zoom, Google Meet, or Microsoft Teams, your sound card mixes incoming speaker packets and microphone audio into a hardware buffer.
Native Loopback Pipelines
Maple's native Rust daemon attaches to CoreAudio on macOS and WASAPI on Windows, streaming raw PCM audio chunks directly into an embedded, quantized Whisper model running in your local RAM. Everything happens locally with sub-15ms processing latency.