SuperAPI AI Directory

xiaoyuzhou-transcript-summarizer

Writing & research

View repository

Updated 10-16 04:41

About this skill

Xiaoyuzhou Podcast Transcript Extraction and Summarization Skill

Workflow

  1. Validate whether the user input is a valid Xiaoyuzhou single-episode link.
  2. Extract the audio URL from the publicly accessible Xiaoyuzhou single-episode page; do not bypass login, paywalls, or access controls.
  3. Download the audio to the local task directory.
  4. Use ffmpeg to split into 30-minute segments.
  5. Transcribe each segment in parallel, and keep the transcript for each segment.
  6. Merge the complete transcript in ascending order by chunk index.
  7. MUSTMust read prompts/summarize_takeaways.md, and strictly follow the requirements of that prompt to summarize the merged transcripts/full_transcript.txt to generate summary.md. Do not switch to temporary ad hoc rules, manual brief bullet points, heuristic extraction, or any other summarization format that does not follow that prompt.

Recommended command

python scripts/run.py "<xiaoyuzhou_episode_url>" --chunk-minutes 30 --workers 4

Optional parameters:

python scripts/run.py "<url>" \
  --output-dir output \
  --chunk-minutes 30 \
  --workers 4 \
  --transcriber faster-whisper \
  --summary-model agent

Output

Each run writes to the project root directory under output/{timestamp}_{episode_id}/。output/ Must be created in the user's current project root directory, not in the Skill directory or any other temporary directory.

  • metadata.json
  • pipeline.log
  • audio/original.*
  • audio/chunks/chunk_000.*
  • transcripts/chunk_000.txt
  • transcripts/full_transcript.txt
  • summary.md

Finally, return to the user the full transcript path, the summary file path, and the summary content. On failure, explain the specific step and the files that have already been generated.

Constraints

  • Only process https://www.xiaoyuzhoufm.com/episode/... or same-domain http links.
  • Do not fabricate audio links; if extraction fails due to page structure changes, report a clear error directly.
  • Do not bypass login walls, paywalls, regional restrictions, or other access controls.
  • Do not fabricate a transcript when transcription fails.
  • The summary must be based on the complete transcript.