forrestahha/dsh-voice-input
Voice-to-text input plugin for the DeepSeek Harness Web UI
Voice-to-text input for the DeepSeek Harness Web UI: adds a microphone button immediately before the composer send button. Uses the browser's Web Speech API, streams recognition results into the current draft, and never submits the message automatically. Supports standard and WebKit-prefixed SpeechRecognition implementations, uses the browser language (falling back to zh-CN), and preserves natural spacing for Chinese, Japanese, Korean, and Latin text. Clicking again stops recording. Unsupported browsers show a disabled microphone button instead of failing at startup.
Install
dsh plugin --profile web add github:forrestahha/dsh-voice-input && dsh webInstall directly from GitHub into the Web profile: dsh plugin --profile web add github:forrestahha/dsh-voice-input, then dsh web. Pinned: dsh plugin --profile web add github:forrestahha/dsh-voice-input#v0.1.1. Requires SpeechRecognition or webkitSpeechRecognition (Chromium-based browsers best). The plugin does not store audio and adds no network calls of its own.
Compatibility
DeepSeek Harness web profile (browser). Requires Web Speech API support; localhost is treated as a secure context, otherwise use HTTPS for mic access.
Details
- Repo: forrestahha/dsh-voice-input
- Category: Web UI Enhancements
- Stars: 3
- Version: GitHub source v0.1.1 (no npm package)
- Last push: 2026-08-14
- First seen: 2026-08-14
Recent updates
mic button before composer; Web Speech API streaming dictation; no auto-submit; browser-language detection; CJK spacing; graceful unsupported-browser fallback.
FAQ
- Which browsers work?
- Any with SpeechRecognition or webkitSpeechRecognition — Chromium-based browsers provide the broadest support; unsupported browsers show a disabled button.
- Does it submit the message automatically?
- No — it streams recognition into the current draft and never sends the message on its own.
- Is audio stored or sent by the plugin?
- No — the plugin stores no audio and adds no network calls; the browser's own speech service handles recognition.
Alternatives
Jesse-njx/dsh-voice · SaiSenBox/dsh-prompt-manager · Scorp1o117/dsh-soul-md