ReelTranscriber

A Python pipeline designed for qualitative social science research. It automatically scrapes audio from short-form social media videos (Instagram Reels, TikTok, YouTube Shorts), runs hardware-accelerated speech-to-text recognition via OpenAI Whisper, and exports sentence-indexed transcripts formatted into a PDF audit report.

Features

• Automated Video Scraper: Ingests direct URLs from Instagram, TikTok, and YouTube via yt-dlp.
• High-Accuracy GPU Transcription: Powered by OpenAI's Whisper model (medium model by default).
• Sentence Indexing ([S1], [S2]): Automatically splits raw speech into discrete, numbered sentence units suitable for qualitative manual coding in software like NVivo or MAXQDA.
• Publication-Ready PDF Export: Generates a PDF report containing video metadata, verbatim transcripts, and sentence index breakdowns.

Key Use Cases

ReelTranscriber is designed to streamline workflows across research, content creation, and data analysis:

• Qualitative Academic Research: Instantly convert large batches of short-form social media videos (Instagram Reels, TikTok, YouTube Shorts) into sentence-indexed transcripts ([S1], [S2]) for deductive content analysis, thematic coding, or manual ingestion into qualitative software like NVivo and MAXQDA.
• Content Creator Auditing: Extract spoken audio from high-performing viral Reels to analyze script structures, hook pacing, and engagement calls-to-action (CTAs).
• Dataset Generation for NLP: Generate structured, sentence-segmented text datasets from spoken social media audio for natural language processing, sentiment analysis, or topic modeling tasks.
• Accessibility & Repurposing: Quickly convert video speech into structured text documents to produce video captions, blog summaries, or downloadable PDF show notes.

---

Getting Started

Prerequisites

To run this pipeline locally or in Google Colab, ensure you have:

• Python 3.9+
• An active GPU environment (recommended: NVIDIA T4 or higher)
• System package: ffmpeg (installed automatically in Google Colab)

Dependencies

Install required Python packages:

pip install openai-whisper yt-dlp reportlab torch pandas openpyxl