README.md

metadata

title: TutorialMaker
emoji: 💻
colorFrom: indigo
colorTo: purple
sdk: gradio
sdk_version: 6.19.0
python_version: '3.13'
app_file: app.py
pinned: false
license: mit
short_description: Make Tutorials from YouTube Videos

YouTube → Tutorial Post Generator

Give a topic and your Hugging Face token, and this Space builds a downloadable .docx tutorial (text + AI-captioned screenshots) from the single best YouTube video on that topic.

Pipeline

1. Search — top 5 videos via the adarshajay/youtube-search Space.
2. Sentiment rank — fetch top comments per video with yt-dlp and score them with the BERT classifier OmarMedhat7/youtube-sentiment-analysis-model; the video with the highest positive share wins.
3. Download the winner with yt-dlp and extract audio with ffmpeg.
4. Transcribe locally with faster-whisper (segment timestamps).
5. Candidate frames are extracted densely from the video, then the video is auto-deleted — only small JPEGs and the transcript remain.
6. Tutorial text — deepseek-ai/DeepSeek-V3 (HF Inference Providers, billed to your token) turns the transcript into structured steps.
7. Screenshot selection — a weighted indicator blends each step's LLM-suggested timestamp with Whisper's actual speech timing.
8. Captions — a vision model (default Qwen/Qwen2.5-VL-72B-Instruct, billed to your token) captions each screenshot.
9. Assemble the .docx for download.

Notes

• Your HF token is used only for the LLM and vision-model calls and is billed to your account. Create a fine-grained token with "Make calls to Inference Providers" at https://huggingface.co/settings/tokens.
• Free CPU tier: Whisper runs on CPU, so transcription is slow — keep videos short (default cap ~20 min).
• YouTube may block the Space's IP. If downloads/comments fail, add a Netscape-format cookie file as a Space secret named YT_COOKIES (the file's contents).

Local run

pip install -r requirements.txt   # needs ffmpeg on PATH
python app.py