mcpField
X
xiaoyuzhou-transcribe

xiaoyuzhou-transcribe

github.com/weifenghuangCreative & media148 installsUpdated 42w ago
Repository

About xiaoyuzhou-transcribe

xiaoyuzhou-transcribe is a script-based skill for creating SRT and TXT transcripts from Xiaoyuzhou episode audio. It suits listeners who need locally generated subtitles, using faster-whisper for speech-to-text.

Key features

  1. 01Download episode audio with resume support
  2. 02Transcribe with faster-whisper locally
  3. 03Export SRT and TXT files
  4. 04Use a local audio file instead
  5. 05Configure model, language, and VAD

Use cases

  1. 01Create subtitles for a Xiaoyuzhou episode
  2. 02Transcribe an audio file already on disk

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/weifenghuang/skillvault --skill xiaoyuzhou-transcribe

After installing, type/xiaoyuzhou-transcribein Claude Code to confirm it loaded.

Bundled filesscripts/xiaoyuzhou_transcribe.py

Details

From the project README

Xiaoyuzhou Transcribe

Overview

Generate SRT and TXT transcripts from a Xiaoyuzhou episode URL by downloading the audio and running faster-whisper locally. Use the bundled script to keep the workflow deterministic and repeatable.

Quick Start

  • Install dependency: python3 -m pip install -U faster-whisper
  • Run: python3 scripts/xiaoyuzhou_transcribe.py "<episode-url>" --output-dir .
  • Expect outputs: xiaoyuzhou-<eid>.mp3, xiaoyuzhou-<eid>.srt, xiaoyuzhou-<eid>.txt

Workflow

  • Fetch the episode page and parse __NEXT_DATA__ to locate the audio URL.
  • Download the audio (resume supported) unless --audio-path is provided.
  • Transcribe audio with faster-whisper and write SRT + TXT.

Script Usage

scripts/xiaoyuzhou_transcribe.py accepts:

  • --output-dir: write outputs to a specific directory
  • --model: whisper model size (tiny by default; use base or small for higher accuracy)
  • --language: force a language code or allow auto-detect
  • --audio-path: transcribe a local audio file instead of downloading
  • --force-download: re-download the audio even if it exists
  • --no-vad: disable VAD filtering if segments are too aggressive

Notes

  • If the episode is private or requires login, the script cannot access it.
  • Xiaoyuzhou does not expose public transcripts for many episodes; this workflow generates subtitles via speech-to-text instead.

Comments

Markdown supported