Add video document type (10_video) with Whisper transcription and OpenCV frame extraction

[Copilot]

Adds video as the 10th document type in the parsing guide, enabling text extraction from video files for RAG pipelines via two methods: audio transcription and keyframe extraction.

New: unstructured_documents/10_video/

• 01_whisper_transcription.py — Extracts audio via ffmpeg, transcribes with OpenAI Whisper, returns timestamped segments suitable for time-indexed RAG chunks
• 02_frame_extraction.py — Interval-based and scene-change keyframe extraction via OpenCV, with build_frame_descriptions() for generating embeddable text from visual content
• sample_docs/generate_samples.py — Generates lecture.mp4 (10-frame ML lecture slides) and short_clip.mp4 (3-frame quick test) using OpenCV

Both scripts follow existing conventions: lazy imports, availability checks with install instructions, graceful degradation, shared chunking integration, and demo functions.

Updated

• pyproject.toml — video optional dependency group (openai-whisper, opencv-python-headless)
• README.md / unstructured_documents/README.md — Document type tables, counts, decision matrix, quick start, repository structure

Usage

uv sync --extra video
uv run python unstructured_documents/10_video/sample_docs/generate_samples.py
uv run python unstructured_documents/10_video/02_frame_extraction.py
uv run python unstructured_documents/10_video/01_whisper_transcription.py

from unstructured_documents.shared.chunking import chunk_by_sentences

# Transcribe and chunk for RAG
result = transcribe_video(Path("lecture.mp4"), model_size="base")
chunks = chunk_by_sentences(result["text"], sentences_per_chunk=5)

# Extract keyframes for visual context
keyframes = extract_keyframes(Path("lecture.mp4"), threshold=30.0)
descriptions = build_frame_descriptions(keyframes)

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.