gettr-transcribe-summarize
от kevin37li
Speech & TranscriptionКратко
Download audio from a GETTR post.
Описание
Что делает этот навык
Downloads audio from a GETTR post URL, transcribes it locally with MLX Whisper on Apple Silicon using VTT timestamps, and produces a bullet-point summary with an optional timestamped outline. Supports both /post/ and /streaming/ URL types, using different extraction methods for each. Does not translate content or handle authentication-gated posts.
Transcription runs fully on-device with MLX Whisper, so no audio is sent to external APIs.
Когда его использовать
- Transcribing a politician's GETTR video post into searchable text
- Creating a timestamped outline of a long GETTR livestream
- Summarizing a GETTR post recorded in Chinese or another non-English language
- Archiving spoken content from GETTR posts as structured notes
- Generating bullet-point takeaways from a GETTR streaming video
Пример сценария
User asks: Transcribe and summarize https://gettr.com/post/p1abc2def with a medium bullet summary
- 1
Parse slug 'p1abc2def' from the URL and set output directory to out/gettr-transcribe-summarize/p1abc2def/
- 2
Run extract_gettr_og_video.py to get the HLS video URL from the post HTML
- 3
Run run_pipeline.sh with the video URL and slug to download 16kHz mono WAV and transcribe via MLX Whisper
- 4
Read the generated VTT file and group adjacent cues into coherent sections
- 5
Write 8-20 bullets and a timestamped outline to summary.md
summary.md containing a bullet-point summary and timestamped outline of the post content
Требования
- mlx-whisper Python package (Apple Silicon Mac required)
- ffmpeg
Установка
clawhub install kevin37li/gettr-transcribe-summarizeВыполните команду CLI или вставьте промпт в Claude Code, Cursor, Codex — любой кодинг-агент, которым пользуетесь.