Voice plugin for DeepSeek Harness: local streaming speech recognition into editable drafts, sentence-by-sentence read-aloud with captions, and voice interruption. Edge online or local VITS/Kokoro TTS; host compatibility is bounded by engines.dsh.

Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.

Persistent voice conversations for DSH with cloud speech recognition, Edge TTS, and background Agent delegation