Claude Couldn’t Watch Videos. One Developer Fixed that With a Clever Trick — Here’s How It Works

2026-07-08 · Source: Towards AI - Medium · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, quick

Summary

Developer Brad Bonanno has released "/watch", an open-source skill designed to overcome a significant limitation in AI tools like Claude. Traditionally, when given a YouTube link, Claude could only infer content from the video's title or an incomplete transcript, missing up to 90% of visual information such as on-screen code or UI interactions. Bonanno's "/watch" tool, available under an MIT license and boasting approximately 2,500 GitHub stars, addresses this by converting video content into a format AI can process. This clever solution allows AI to effectively "watch" videos, enabling a deeper understanding of their visual elements rather than relying solely on textual metadata.

Key takeaway

For AI Engineers integrating large language models with multimedia, Brad Bonanno's "/watch" skill offers a critical solution for video content analysis. You should consider implementing this open-source tool to enable your AI applications, like Claude, to process visual information from YouTube links, moving beyond limited text-only interpretations. This enhances AI's utility for tasks requiring detailed visual context, such as debugging or content summarization.

Key insights

An open-source tool enables AI to process YouTube video content by converting visual information into an AI-readable format, overcoming transcript limitations.

Principles

• AI's visual processing of video remains a significant challenge.
• Open-source tools can rapidly bridge AI capability gaps.
• Transcripts alone are insufficient for video content analysis.

Method

The "/watch" skill processes video by slicing it into AI-perceivable components, enabling AI models to interpret visual information beyond titles or transcripts.

In practice

• Analyze on-screen code in technical tutorials.
• Identify specific UI bugs from video recordings.
• Understand visual context of viral video clips.

Topics

• AI Video Analysis
• Claude AI
• Open-source Tools
• YouTube Integration
• Visual Content Processing

Best for: AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• Can you *really* train AI to "get" videos just by showing it a million of them? — Current video models excel at generation but lack systematic evaluation for genuine spatiotemporal reasoning.
• How I Use AI To Run My YouTube Channel — Integrating YouTube analytics with an AI chatbot can streamline content strategy and production.
• I compared how Gemini, ChatGPT, and Claude can analyze videos - this model wins — Gemini Pro excels at native video understanding, while ChatGPT requires external tools like Codex for similar capabilities.
• 😺 🎙️ We're LIVE now & NEW LTX interview drop — AI agent and video generation tools are rapidly advancing, enabling local, high-quality creative production.
• How SEO Teams Are Turning Claude Into a Full Content Pipeline with Walter MCP — Integrating AI content tools into a single pipeline streamlines post-draft workflows, ensuring quality and consistency at scale.
• What is AI Video Editing and How Does It Work? — AI video editing automates and enhances video production through machine learning, computer vision, NLP, and audio analysis.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Towards AI - Medium.