Skip to main content
H

dsh-read-aloud-edge

hexiao160705-dev/dsh-read-aloud-edge

Speaker button on every assistant reply that synthesizes the reply with the locally installed edge-tts and plays it in a small player with 1x-4x speed control; audio is cached by content hash and the reply text never leaves the machine to a paid API.

Install

dsh plugin --profile web add github:hexiao160705-dev/dsh-read-aloud-edge

README

dsh-read-aloud-edge

Read assistant replies aloud in DeepSeek Harness (DSH), with 1×–4× playback speed.

Renamed from dsh-read-aloud: that name is already taken on npm by a different plugin. This package is the Edge TTS edition — it synthesizes the reply locally with edge-tts and plays it with speed control.

DSH ships speech-to-text (voice input) but no text-to-speech model. This plugin uses the locally installed edge-tts (Microsoft Edge online voices) to synthesize the reply, then plays it in a small player with speed control.

Features

  • Speaker button in every assistant reply's action row
  • Reads only the reply text, not tool cards or reasoning
  • Player with play/pause, progress, and 1× / 1.25× / 1.5× / 2× / 2.5× / 3× / 4×
  • Pitch is preserved when speeding up (preservesPitch)
  • Audio is cached under ~/.dsh/read-aloud-edge/cache/ by a content hash
  • Chinese text uses zh-CN-XiaoxiaoNeural; English text uses en-US-AriaNeural

Requirements

  • edge-tts installed and on one of the known paths, or set EDGE_TTS_PATH
  • Network access to the Edge TTS service (the reply text is sent to Microsoft's TTS endpoint)

Install:

pip install edge-tts
# or
uv tool install edge-tts

Install the plugin

dsh plugin --profile desktop add github:hexiao160705-dev/dsh-read-aloud-edge

Or from a checkout:

dsh plugin --profile desktop add /absolute/path/to/dsh-read-aloud-edge

Reload/restart DSH. A speaker icon appears in the action row of each assistant reply.

Privacy

  • The host reads the text from the request, synthesizes it through edge-tts, and returns only a cache id + audio URL.
  • No API key, account id, or token is read or sent.
  • edge-tts sends the text to Microsoft's online TTS service; the audio is cached locally. If the reply must not leave the machine, do not use this plugin (a local TTS engine would be needed).
  • Cache files stay under ~/.dsh/read-aloud-edge/cache/; delete that directory to clear them.

Limitations

  • Edge TTS is an online service and may be unavailable or rate-limited.
  • Very long replies are truncated to 8,000 characters per synthesis.
  • Speed changes use HTML audio playback rate; at 4× the result is fast but pitch-preserved.
  • The reply text is located from the rendered chat row; if DSH changes its DOM attributes, the button falls back to the current text selection.

License

MIT

Related plugins