Video Analysis Python Packages

claude-real-video

Let Claude (or any LLM) actually watch a video — scene-aware, deduplicated frames + transcript, from a URL or local file. Runs locally, MIT.

video-research-mcp

MCP tools and Claude Code workflows for video analysis, cited research, document extraction, knowledge storage, and optional explainer video production.

watch-skill

Give AI agents eyes, ears, and verifiable results. Watch Skill turns video, audio and screen activity into searchable, timestamped evidence and proves work with deterministic contracts, not model opinion. DeepWatch is the agent workspace built on DeepSeek Harness. Python + npm, MCP, CLI, REST, Web.

yueying

Let AI watch videos: local files or YouTube/Bilibili URLs -> timestamped transcript + keyframes + contact sheets. Offline, no API key. MCP server + CLI + agent skill.

talkthrough-mcp

Recordings in. Agent-ready evidence out. Local-first MCP for transcripts, keyframes, OCR, speakers, and wall-clock evidence.

video-agent-mcp

MCP tools and Claude Code workflows for video analysis, cited research, document extraction, knowledge storage, and optional explainer video production.

video-explainer-mcp

MCP tools and Claude Code workflows for video analysis, cited research, document extraction, knowledge storage, and optional explainer video production.

sportvision

Real-time sports analytics toolkit built on the Roboflow ecosystem. Detects players and ball via RF-DETR, tracks with ByteTrack, classifies teams by jersey color, and computes possession, speed, distance, and heatmaps.

shotdrift

Measure whether a video holds the camera move it was given. Camera path from pixels alone - pan, zoom, roll - with thresholds calibrated against real footage.

kodokan

Study Kodokan Judo throws from video via body-pose analysis

tilia-library

A GUI for graphical analysis and annotation of video and audio files.

tilia-core

A GUI for graphical analysis and annotation of video and audio files.

video-vision-mcp

MCP server that turns any video (file, URL, or Jira attachment) into frames + transcript for Claude Code

frame-ingest

Claude Code skill that watches a video (local file or YouTube/URL link) and writes a citable Markdown document: timestamped transcript, chapters, summaries, verbatim quotes. No API key needed. Also works in Codex and other agents; includes a CLI.

eyeroll

AI eyes that roll through video footage — watch, understand, act

pollux-ai

Multimodal orchestration for LLM APIs. Source patterns, context caching, and structured output for text, PDFs, images, video, and YouTube - so you don't manage the complexity yourself.

vidchain

✅A Lightweight Video RAG Framework for Multimodal Reasoning

analysis-video

Turn a lecture video into AI-readable context: one markdown section per screen, with the frames and the transcript that belong to it. Agent-first CLI.

llm-real-video

Let Claude (or any LLM) actually watch a video — scene-aware, deduplicated frames + transcript, from a URL or local file. Runs locally, MIT.

iflow-mcp-youtube-mcp-server

A comprehensive Model Context Protocol (MCP) server providing real-time YouTube Data API access for AI assistants. Features 14 functions including intelligent content evaluation with technology freshness scoring for knowledge base curation.

video-to-llm

Turn hours of local video into one timestamped, citable document. Process once, ask forever. Fully offline, resumable, any language.

openscenesense

OpenSceneSense is a Python library that harnesses AI for advanced video analysis, offering customizable frame and audio insights for dynamic applications in media, education, and content moderation.

analysis-video-gui

Turn a lecture video into AI-readable context: one markdown section per screen, with the frames and the transcript that belong to it. Agent-first CLI.

loups

Extract video chapter timestamps and title screens using template matching and OCR - perfect for sports, podcasts, and content creation