dsh-voice: DeepSeek Harness Web Voice Message Plugin
`dsh-voice` is a voice plugin for DeepSeek Harness, designed to enhance the interactivity of the Web UI. It leverages the Microsoft Edge TTS engine to enable agents to proactively send voice messages and users to listen to audio. The plugin implements its core features by registering the `voice_say` tool and HTTP routes, and renders a player in the browser. During installation, its path must be added to the configuration and the Profile restarted. The plugin requires no API key, supports multiple neural voices, caches audio files locally, and is suitable for developers needing rich voice interaction scenarios.
Read Moredsh-voice: A Local Voice Input Plugin for DeepSeek Harness
dsh-voice is a DeepSeek Harness plugin that provides local speech-to-text (STT) for the Web UI and TUI. It uses the host-side ONNX-based Whisper service and integrates the microphone in the browser, enabling privacy-safe voice input without external binaries. To install, add it to a DSH Profile; configuration is optional. It supports downloading various Whisper models from Hugging Face. Note that TUI mode depends on system recording tools, while in Web mode audio is recorded locally and transferred over HTTP.
Read More@motong/dsh-voice: Voice Input and Text-to-Speech Plugin for DeepSeek Harness
`@motong/dsh-voice` is a community plugin that provides voice capabilities for DeepSeek Harness (DSH). It requires no external API key and supports speech-to-text (STT, with hotkey invocation) and text-to-speech (TTS, based on Microsoft Edge voices). Users can dictate using the `Ctrl+Shift+Space` shortcut or by clicking a button, and switch between voices in settings. The plugin relies on native browser features and is compatible with Chrome/Edge; TTS requires an internet connection. It is open-sourced under the MIT license.
Read Moredsh-voice: Official Voice Plugin for DeepSeek Harness Web GUI
The DeepSeek Harness (DSH) plugin `dsh-voice` provides voice input for the Web GUI through a modular plugin architecture. It integrates a microphone button and real-time waveform display, and supports three transcription engines, including local Whisper.cpp and the OpenAI API. The recording bar offers three actions: cancel, insert text, or send directly. Recordings longer than 22 seconds are automatically segmented. The plugin does not require the Handy app, supports 67 models, stores data locally, and ensures the security of API keys. Installation is performed via CLI commands and depends on Bash, curl, ffmpeg, and an HTTPS/localhost environment.
Read Moredsh-voice: A Turn-Based Voice Plugin for DeepSeek Harness Interview Scenarios
dsh-voice is a turn-based voice channel plugin for DeepSeek Harness (DSH), designed specifically for interview scenarios. It abstracts underlying engine details through Agent tools, supports independent selection of TTS/ASR engines, and provides speech synthesis, audio transcription, and health check functionality. The plugin supports cloud engines (Qwen, MiMo) and local engines (Sherpa-onnx, Whisper, etc.), with engines and parameters defined via a configuration file after installation. In an interviewer preset, the Agent can be instructed to use `voice_speak` to read questions and call `voice_listen` to wait for responses. The current version is v0.1, follows a turn-based-first strategy, and does not yet support full-duplex interruption or session persistence.
Read Moredsh-voice: Zero-Latency Streaming Voice Input for DeepSeek Harness
The DeepSeek Harness plugin `dsh-voice` is an open-source voice interaction component designed to address manual-input latency and improve handling of complex instructions. It captures audio locally in the browser and uses VAD technology to enable zero-latency streaming dictation, supporting haptic press-to-talk, automatic voice-message sending, and zero-latency subtitle display. With multi-provider fallback chains, offline Whisper transcription, and support for engines such as SenseVoice, it ensures transcription accuracy and stability. The plugin also includes hardware noise reduction, real-time audio streaming, and visualizations, while all API keys are resolved on the host side to comply with security and privacy requirements.
Read Moredsh-voice: DeepSeek Harness Voice Input Plugin
dsh-voice is a voice input plugin for DeepSeek Harness (DSH). It adds a microphone button to the bottom-right corner of the input box and supports one-click recording-to-text. The plugin provides three speech recognition engines: `webspeech`, `xfyun` (iFlytek), and `whisper`, and supports multi-device selection, automatic gain, and stitching of long utterances. Installation requires executing a specific command and restarting the service, with the configuration file located in `client.js`. Using the iFlytek engine requires setting local environment variables or the configuration file. On Chrome in mainland China, `webspeech` may cause errors, and the recognition language for `whisper` must be set manually.
Read Moredsh-voice: DSH TTS and STT Plugin
dsh-voice is a voice processing plugin based on the DSH framework, designed to provide simplified TTS (Text-to-Speech) and STT (Speech-to-Text) capabilities for agents. The plugin adopts the MIT license, requires Node.js version 22 or above, has no runtime dependency on internal modules, and implements communication via native WebSockets. Its core functions consist of three tools: `voice_tts` supports free synthesis of text into MP3 files, allows customization of parameters such as voice timbre and speed, and can be used with zero configuration; `voice_stt` is used for audio-to-text transcription, supports OpenAI-compatible interfaces, and requires configuration of an ASR key (recommended to be managed via environment variables); and `voice_list` is used to query the list of available voices. The plugin is accessed via `dsh plug...
Read Moredsh-voice: Voice input to user message, readable replies.
The DSH plugin `dsh-voice` aims to enhance terminal agent efficiency through voice interaction, implementing speech-to-text, text-to-speech, and reading aloud. Adhering to a "local-first" philosophy, the plugin stores audio files locally by default and does not automatically record or upload to the cloud unless users explicitly configure cloud backends such as OpenAI or Edge-TTS. Its core functions include: `transcribe`, which supports transcription from files or real-time microphone recording, inserting results into the session as voice cards; `speak`, which synthesizes and plays text in the background, suitable for announcing long task status; and the `/voice` command, used to toggle session-level auto-read replies (defaulting to off). The plugin is compatible with Node.js >=20, supports local backends like Whisper-local and macOS Say, and can work with `dsh-cr`.
Read Moredsh-voice: Voice Input and Text-to-Speech Plugin for DeepSeek Harness
dsh-voice is a voice plugin for the DSH Web client, designed to extend the voice interaction capabilities of the DeepSeek chat interface. The plugin converts speech to text on the input side and reads text aloud on the output side, while also supporting audio file transcription and synthesis. Key features include a microphone button in the Web UI for voice input and a read-aloud button for assistant replies (based on browser speechSynthesis); it provides the `voice_transcribe` tool to transcribe audio attachments via Whisper-compatible endpoints and the `voice_speak` tool to synthesize voice files via OpenAI-compatible endpoints. The plugin does not require additional API keys or models and supports configuring backends such as Whisper and Kokoro. Installation is achieved by adding it via a local path.
Read Moredsh-voice: Adding Full-Duplex Voice Mode to DeepSeek Harness
`haoku123/dsh-voice` is a full-duplex voice plugin for DeepSeek Harness (DSH), licensed under MIT, designed to provide real-time voice interaction experiences for web clients. Its core pipeline follows the flow of "streaming ASR → LLM → TTS with interruption support." The browser side captures Raw PCM audio, while the host side leverages `SenseVoice` (based on `sherpa-onnx`) for Chinese speech recognition, punctuation prediction, and text normalization, and uses `msedge-tts` to stream audio and captions back to the client. The plugin supports three input methods: tap-to-talk for continuous dictation, hold-to-record via the send key or microphone (swipe up to discard), and keyboard-free dictation by holding the Ctrl key (Esc to discard). It also features real-time caption preview.
Read More