halo-kokoro · 24 khz mono wav · onnxruntime · local
hear the 1bit monster talk.
Type a line, pick a voice, hit synthesize. The audio comes from your halo box over same-origin /kokoro/v1/audio/speech. No cloud. No telemetry. One GET, one WAV back.
ready · same-origin POST /kokoro/v1/audio/speech
Endpoint shape mirrors OpenAI's /v1/audio/speech. If you're scripting, POST JSON {"model":"kokoro","input":"...","voice":"af_sky","format":"wav"} and you'll get audio/wav back.