dsh-ears

When using DeepSeek Harness for daily conversations or content creation, relying solely on text input can sometimes limit your expressive efficiency. The dsh-ears plugin is designed to solve this exact pain point by giving a pair of ears to the text-only interface, allowing users to effortlessly dictate their thoughts. Whether you are quickly capturing fleeting inspirations or engaging in long-form dictation, this plugin significantly enhances the input experience, making interactions more natural and fluid.

In terms of core capabilities, dsh-ears offers an exceptionally rich selection of speech recognition backends. You can use the browser's native Web Speech for real-time recognition, run a local Whisper model to ensure privacy, or integrate with Groq, Deepgram, Alibaba Cloud, Tencent Cloud, and any OpenAI-compatible API. All API credentials are securely stored on the Host side. Additionally, the plugin features built-in LLM-powered transcription polishing that automatically removes filler words, corrects typos, and handles self-corrections. All transcriptions are safely written to an editable draft and never sent automatically, giving you complete control over the final content.

This plugin is perfectly suited for DeepSeek Harness users, content creators, and developers who demand highly efficient input methods. If you prefer thinking out loud or need the flexibility to switch between cloud and local speech recognition solutions across different scenarios, dsh-ears will be an indispensable productivity tool for your workflow.