finds.dev

JimLiu/whisper-subtitles

Apple PodCast Transcription with OpenAI's Whisper

View on GitHub

★ 345Jupyter NotebookMITupdated Dec 2023

// the take

A grab-bag of Jupyter notebooks for running Whisper and WhisperX against podcasts, YouTube videos, and Bilibili clips to generate subtitles, including a variant with speaker diarization. It's for someone who wants a copy-paste Colab-style notebook to transcribe a specific episode or video, not a library to build on.

Covers several real transcription sources (Apple Podcasts via RSS/download, YouTube, Bilibili) rather than just a toy example. The speaker-diarization notebook and the WhisperX variants (faster inference, word-level timestamps) give you options beyond vanilla Whisper depending on what you need. Notebooks are a low-friction way to try this without setting up a project.

No code outside the notebooks themselves — no shared module, no tests, no packaging, so anything beyond the exact scenario in a given notebook means editing cells by hand. The README is a single sentence and doesn't document setup, model size tradeoffs, or how podcast URLs are resolved to audio, so you're reading notebook cells to figure out usage. No activity since December 2023, and Whisper/WhisperX/yt-dlp (which this likely depends on for YouTube) all move fast enough that a two-year-old pin is a real risk of broken dependencies or dead extraction logic. Nothing here handles longer-form content efficiently — no chunking/batching strategy mentioned, so multi-hour podcasts will be slow and memory-heavy on consumer GPUs.

We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →

// at a glance

Stars
345
Forks
42
Language
Jupyter Notebook
License
MIT
Last commit
3 Dec 2023

// nearby

If whisper-subtitles isn't quite it.

All repos →