dsh-voice
Run the following command in DeepSeek Harness:
dsh plugin install haoku123/dsh-voice
Paste the following prompt into your AI chat to install this plugin:
To install this plugin in DeepSeek Harness, run the command dsh plugin install haoku123/dsh-voice, with the full source address at https://github.com/haoku123/dsh-voice.
About this plugin
The dsh-voice plugin addresses the challenges of latency and interruption in voice interaction, bringing a full-duplex voice mode to DeepSeek Harness. Unlike traditional voice assistants that wait for silence before responding, it employs a streamed ASR → LLM → TTS architecture to enable near-real-time conversation: users can speak while the system listens, processes, and generates spoken replies, with true barge-in support that lets the user interrupt and redirect the assistant at any moment. Key capabilities include press-to-talk dictation, live captions, SenseVoice-powered Chinese ASR with punctuation and ITN, and a three-layer interruption detection system that clears playback queues, bumps synthesis epochs, and cancels running turns. Performance is optimized through a cache-through proxy and incremental decoding, ensuring smooth multi-turn voice exchanges. Ideal for developers building AI applications, enthusiasts integrating voice features, or anyone needing high-quality Chinese voice interaction with DeepSeek Harness.
Screenshots
Use Cases
- Enable real-time interaction and interruption in voice assistant dialogues.
- Add voice input/output capabilities to AI applications.
- Create streamed voice reply experiences with live captions.
Best For
- Voice application developers and integrators.
- AI builders needing optimized Chinese voice capabilities.
- AI enthusiasts and interaction design researchers.
Related Plugins
ModLens is a vision plugin for DeepSeek Harness that gives text-only models sight by reading images pasted directly into chat, with zero-config setup and multiple vision engines.
On-demand vision for text-only DeepSeek Harness agents: built-in free keyless vision chain and 14 vision tools, routing image turns as tool calls to vision models with pixel fidelity, no Python needed, one-command install.
Give text-only models in DeepSeek Harness eyes, enabling image Q&A, long-screenshot OCR, UI restoration, and GUI visual tasks.
An armor-breaking plugin for DeepSeek: rewrites instruction priority, expands refusal-blocked phrases, persists across sessions, and shows a green active indicator.