Voice & Audio

What it is — This is a DeepSeek Harness voice reading plugin that uses Doubao TTS natural timbre to point-read AI replies.

Who it is for — If you want to use natural timbre to read AI replies in the DeepSeek Harness desktop with a macOS Omi engine, you can install this plugin. Windows users can skip it because it does not support Windows.

Watch out — Sandbox test passed: it was installed in a fresh profile and registered with harness. No obvious pitfalls were found. It requires configuring the Doubao API Key in the Omi engine and users bear the per-character cost.

The verdict — I would install it if I am using DeepSeek Harness on macOS with Omi engine for natural voice reading, because it supports point read, pause and continue with text filtering, but I would not if no macOS or preferring zero-cost mechanical voice.

What it is — dsh-ears provides voice transcription and optional polishing for DeepSeek Harness, supporting multiple backends and outputting editable drafts.

Who it is for — It suits users performing text tasks in the browser environment. Users primarily using keyboard input may not need it.

Watch out — A refresh of the Web UI is required after installation to display the microphone icon. Sandbox testing confirmed functionality in a new profile. No obvious issues were found.

The verdict — I will install it because it allows retaining the original transcription and manual editing of drafts.

What it is — Integrates Edge TTS and RVC for voice reading of assistant replies in DeepSeek Harness web UI.

Who it is for — In the DeepSeek Harness web profile, this plugin enables voice reading of assistant replies. Users who do not use the web profile will not be able to use this plugin.

Watch out — Sandbox test confirmed successful installation in a fresh profile and registration by Harness. No obvious issues found.

The verdict — I would install it because the sandbox test passed and it supports RVC custom voices, but only if voice feedback for replies is needed, with the premise of using the web profile.

What it is — dsh-speak plugin for DeepSeek Harness, announces final reply via voice.

Who it is for — When using DeepSeek Harness for long coding tasks and wanting voice prompt on completion. Users who prefer text-only logs without voice features do not need it.

Watch out — Sandbox test passed: installed in fresh profile and registered by harness. No obvious pitfalls found. The project stresses best-effort execution that will not interrupt sessions.

The verdict — I would install it because it offers a verified solution for voice announcements on long tasks.

What it is — DeepSeek Harness voice input plugin supports speaking into the microphone and submitting recognized text as a normal chat message via local or browser speech recognition.

Who it is for — If using DeepSeek Harness for conversations where quick input is needed, you can install it. If using DeepSeek Harness for pure keyboard input, you need not install it because they may have other input methods available.

Watch out — 实测通过，在全新 profile 中安装并被 harness 注册进 profile。没发现明显的坑。

The verdict — If I need microphone input for chatting, I would install it because it provides an alternative input method without affecting the persona.

What it is — Integrate voice input into the Composer toolbar of DeepSeek Harness Web UI, supporting tap-to-monitor live transcription and hold-to-talk voice chat.

Who it is for — When using DeepSeek models for conversation in the DeepSeek Harness Web UI and needing voice input support, install this plugin. Users who rely primarily on keyboard input won't need it.

Watch out — Sandbox test passed: installed successfully in a fresh profile and registered with harness. No obvious issues found. Requires the host Edge TTS capability for natural reply reading.

The verdict — I would install it because it provides voice input for the composer, but only worth it if you need real-time voice interaction.

What it is — This plugin rings your phone via CallKit, provides call_me and text_me tools, and supports turn-end and approval calls with spoken answers transcribed back into the session.

Who it is for — When running long tasks that need manual confirmation or voice replies, this plugin fits with DSH. Users without need for manual confirmation do not need it, because they can rely on other input methods to finish tasks.

Watch out — Sandbox testing passed: it registers in a new profile via harness. No obvious issues found.

The verdict — I would install it for agents needing phone-based turn-ends, but only if phone call delays are acceptable.

What it is — Perlica-themed DeepSeek Harness plugin delivering tiered terminal notification sounds for task states.

Who it is for — Users performing agent tasks with DeepSeek Harness are suitable for installing this plugin. Those who prefer complete silence can skip it because they may focus more on text-based interactions.

Watch out — Sandbox test passed: installed into a new profile and registered by Harness. It executes shell commands to play sounds. No obvious issues found.

The verdict — I would install it for agent workflows needing state feedback, because it offers targeted audio cues, yet only if background playback is allowed on my platform.

What it is — dsh-ding notifies you when the conversation finishes by playing a sound and showing a Windows notification.

Who it is for — When you use DeepSeek Harness to handle agent tasks that require attention after completion, it is suitable to install in your profile. Users who only view conversations in the browser without needing notifications do not need to install it.

Watch out — Manual approval of the build script is required when installing from source. Sandbox test passed with no obvious issues found.

The verdict — I would install it because it provides notifications when the agent returns to idle state.

What it is — Notification plugin: agent proactively pushes toast desktop notifications, Chinese TTS voice and sound effects, with 60s confirmation window and voice callback.

Who it is for — When the agent uses the notify_user tool to proactively report task status or errors, this plugin is suitable for installation. If not running on Windows with PowerShell 5.1+, users may not use voice features.

Watch out — Sandbox test passed: installed in a new profile and registered in harness profile. No obvious issues found.

The verdict — I would install it because it supports agent proactive voice callback reminders, but only on Windows + PowerShell 5.1+.

What it is — This plugin adds voice input to DeepSeek Harness web GUI's composer, letting you click the microphone to transcribe speech into text.

Who it is for — If you are using DeepSeek Harness web GUI to chat with an LLM and want to input text by speaking, especially with the browser local engine on Chrome. Users who prefer keyboard input without needing voice conversion or mic buttons can use other methods.

Watch out — Sandbox testing passed: the plugin was installed in a fresh profile and registered into the profile by Harness. No obvious issues found.

The verdict — I would install it because it provides a zero-config speech-to-text option using the browser local engine on Chrome or Edge.

What it is — Adds microphone voice input to the DeepSeek Harness composer, supporting live transcription with browser Web Speech API.

Who it is for — Suitable for users needing microphone voice input in the DeepSeek Harness Web UI composer. If not using a browser that supports Web Speech API or not running in browser environment, users do not need to install this plugin.

Watch out — Sandbox test passed after installation in a fresh profile and was registered into the profile by harness. No obvious issues found.

The verdict — I would install it because it deduplicates repeated text and auto-continues long sentences with zero server components via browser Web Speech API.

What it is — Integrates uisfx sound effects into DeepSeek Harness for semantic task and button cues.

Who it is for — For users executing AI agent tasks in DeepSeek Harness who need sound effects for task status and button interactions. For users who do not perform any agent tasks, this plugin is unnecessary.

Watch out — Sandbox test passed, the plugin was installed in a new profile and registered to harness. No static check issues found.

The verdict — I would install it because it exposes the ctx.uisfx service API for integration by other plugins, but only when using dsh version 0.1.0-rc.6.

What it is — Integrates full-duplex voice mode into DeepSeek Harness Web UI.

Who it is for — Users performing Chinese voice conversations in DeepSeek Harness Web UI who require SenseVoice ASR transcription should install this plugin. Users using non-Chinese language models can choose other plugins.

Watch out — No obvious issues found. Sandbox tests passed. Installation requires Node.js 22.19 or higher and from source install need to manually allow the build script.

The verdict — I would install it because it provides true barge-in without API key, but only if using compatible voice models.

What it is — dsh-voice provides transcribe and speak tools in DeepSeek Harness for voice notes as user messages and spoken agent replies.

Who it is for — When handling terminal tasks that require dictating audio notes or having the agent narrate replies aloud using DeepSeek Harness, installing dsh-voice is suitable. Users who do not need audio-based interactions should skip it.

Watch out — Sandbox testing reveals it installs but its build script is blocked by pnpm targeting github:Jesse-njx/dsh-voice, requiring manual allowance for proper registration. No other obvious issues found.

The verdict — I would install it because it enables local-first voice features, but only after manually allowing the build script.

What it is — This plugin adds a microphone button to the right of the chat input box in DeepSeek Harness Web UI, driving local VocoType offline speech recognition and automatically inserting recognized text into the input box.

Who it is for — Users who frequently need to input Chinese text via voice into the chat interface in DeepSeek Harness Web UI are suitable for installing this plugin. Those who mainly input in desktop applications or do not rely on real-time speech functions can use other input methods.

Watch out — Installation requires placing the plugin in the profile packages directory and running pnpm add file:, then adding a mount line in cordis.patch.yml and restarting the interface. The first run of VocoType requires downloading approximately 25MB of installer and 1.6GB of model files. It was tested to register normally in a fresh profile.

The verdict — If you need offline Chinese speech input in DeepSeek Harness Web UI and have sufficient storage, it is worth installing; otherwise, the time spent downloading the model is not worthwhile.

What it is — Enables agents to initiate voice calls to humans using the offer_call tool.

Who it is for — When an agent uses the offer_call tool to actively contact humans. Not suitable for users who rely solely on text-based tool calls without voice capabilities.

Watch out — Requires cordis.patch.yml configuration with CrispASR paths and voice settings. Sandbox testing confirmed successful registration in a new profile. No obvious issues found.

The verdict — I would install it if the agent needs to actively communicate via voice, but only if local speech engines are supported; otherwise the setup effort is not justified.

What it is — dsh-voice is a DeepSeek Harness plugin that provides edge-tts speech synthesis and OpenAI-compatible ASR transcription.

Who it is for — When building a conversational agent with DeepSeek Harness that needs to perform voice dialogue, this plugin can be installed. In projects that only involve text interaction without voice needs, it can be skipped.

Watch out — Sandbox test passed, successfully installed and registered in a fresh profile. No obvious pitfalls found. It executes shell commands.

The verdict — I would install it because it provides free edge-tts speech synthesis capability.

What it is — DeepSeek Harness Web UI plugin for per-workspace completion ringtones and attention sounds.

Who it is for — DeepSeek Harness Web UI users managing tasks needing per-workspace ringtones and human intervention alerts. Users without bundle-plugin support in DSH can skip this plugin.

Watch out — Sandbox test passed in a new profile. Requires DSH bundle-plugin support and pnpm on PATH. No obvious issues found.

The verdict — I would install it because it adds per-workspace ringtones and human event alerts to DSH web UI, but only in DSH versions with bundle support and pnpm available.