This archive holds machine-generated transcripts and summaries for the Shell Game podcast episodes, together with the pipeline that produces them. The pipeline runs as 5 numbered stages; main.py executes them in order.
Shell-Game-Transcripts/
├── docs/
│ ├── README.md this page
│ ├── quickstart.md shortest path to one finished episode
│ ├── installation.md prerequisites, GPU build, dependencies
│ ├── configuration.md the hardcoded knobs and where they live
│ ├── architecture.md data flow, layout, resumability
│ ├── api.md external services this depends on
│ ├── usage.md running it day to day
│ ├── pipeline.md what each numbered stage does
│ ├── faq.md licensing, accuracy, speed
│ ├── troubleshooting.md concrete failures and fixes
│ └── roadmap.md known gaps and non-goals
├── main.py runs every stage in order
├── 1_download.py
├── 2_tagger.py
├── 3_transcriber.py
├── 4_summarizer.py
├── 5_cleanup.py
├── 2024/
├── 2025/
└── requirements.txt
- Quickstart — install, pull a model, transcribe one year
- Installation — prerequisites, GPU acceleration, dependencies
- Configuration — models, language, chunk size, and which file each lives in
- Architecture — data flow, on-disk layout, how resumability works
- External services — what this depends on that it does not control
- Usage — running the pipeline and the helper scripts
- Pipeline — what each stage does and what it writes
- FAQ — licensing, accuracy, speed, regenerating output
- Troubleshooting — concrete failures and their causes
- Roadmap — known limitations and deliberate non-goals
The code is MIT. The transcript text is not — see CONTENT_LICENSE.md.