# Talk to Illustrated Notes > Talk to Illustrated Notes is an AI agent skill for Claude Code, Codex CLI, Gemini CLI, and other agents that read SKILL.md. Turns a YouTube talk into notes that carry the real slides and diagrams inline. Best for: YouTube video to notes AI. - URL: https://atharvashah.com/ai-skills/video-to-illustrated-notes - Price: $49, one payment - Works with: Claude Code, Codex CLI, Gemini CLI, Cursor, GitHub Copilot - Category: Research and knowledge - Built by: Atharva Shah (https://atharvashah.com) - Requirements: Needs a paid Claude plan with Claude Code, plus yt-dlp, ffmpeg, ImageMagick, Python, and Real-ESRGAN. All are free. Upscaling runs on any GPU with Vulkan support, and a slower fallback works without one. Only works on videos that have subtitles. - Buy: https://buy.polar.sh/polar_cl_R6zXD453Fqls2A2MvMzI0peVQkdv2aw1TZP2c1jBqZo?custom_field_data.skill-premium=video-to-illustrated-notes (the skill folder and setup guide arrive by email) ## Turn any recorded talk into notes with the real slides. Paste a YouTube link and get a study note with every slide and diagram placed where the speaker reached it. A transcript summary drops the screen, and this keeps it. ## What it produces - **The illustrated note.** A study note of a recorded talk, with each real slide placed at the point it was shown. ![A real study note from a 95 minute workshop, with the two jobs, create and capture slide set inside the text.](https://atharvashah.com/ai-skills/video-to-illustrated-notes/scene-1.webp) - **Slides in order.** Every slide from the screen share, cropped, readable, and in speaking order. ![Six real slides captured from a recorded marketing workshop, laid out in a numbered grid in the order shown.](https://atharvashah.com/ai-skills/video-to-illustrated-notes/scene-2.webp) - **Point plus diagram.** Each section pairs what the speaker said with the diagram they drew, so you skip the video. ![A real note section titled where to start, showing the speaker's diminishing versus compounding slide above the text.](https://atharvashah.com/ai-skills/video-to-illustrated-notes/scene-3.webp) ## If you can paste a link, you can keep the talk. You install the tools once. Every talk after that goes through the same three steps. 1. **Paste the link.** It checks for subtitles and a working download first. If either is missing, it tells you and stops, instead of handing back a text-only note. 2. **It finds every slide.** It crops to the shared screen, samples frames, removes the repeats, and lays the rest out on contact sheets so it can pick the keepers by eye. 3. **Read the illustrated note.** Slides are upscaled, named in order, and placed at the moment the speaker reaches them, each with a caption and a timestamp. ## The usual options lose the screen. Transcript tools skip the diagrams. Doing it by hand eats the evening. ### Talk to Illustrated Notes - **Slides inline.** Each image sits where the speaker reaches it. - **Captions you can search.** Labels and numbers are written out under each image. - **Faithful upscaling.** Small slide text stays spelled right. - **Source table.** Video, channel, speaker, and date sit at the top. ### Transcript summarizers - **Half the talk is gone.** Tables and charts the speaker never reads aloud vanish. - **Numbers go missing.** The figure on the slide never makes it into the summary. - **No way to check.** You cannot see the slide the summary refers to. - **Generic structure.** Every note reads like the same bullet list. ### Screenshots by hand - **Hours of scrubbing.** You pause, crop, and save one slide at a time. - **Missed changes.** A quick slide flips past while you type. - **Blurry frames.** A 720p screen share stays hard to read. - **Messy filenames.** A folder of image-0142.png means nothing a year later. ## Frequently asked questions ### Can AI turn a YouTube video into notes with screenshots? Yes. This skill pulls the transcript, captures the slides from the video, removes duplicates, and places each slide next to the part of the talk that explains it. ### Which Claude plan do I need? A paid plan with Claude Code, because it downloads video and runs local tools. Pro is enough for one talk at a time. ### What else do I pay for? Nothing. yt-dlp, ffmpeg, ImageMagick, Python, and Real-ESRGAN are free, and every step runs on your machine. ### Does the video need subtitles? Yes. It needs manual or auto-generated captions. Without them it stops and tells you, because a note with no reliable transcript is not worth keeping. ### What kinds of video work best? Talks, workshops, webinars, and conference sessions with a screen share. Freeform canvases and whiteboards work too, with more frames to review. ### Why not use the sharpest upscaler? The sharpest common model invents letter shapes on small text. It looks crisp and misspells words. The default model keeps the text as it was, which matters when you quote a slide. ### Do I need a GPU? A GPU with Vulkan support makes upscaling take a few seconds per frame. Without one, it falls back to a slower resize that still keeps text faithful. ### Which note app does it write for? Any app that reads markdown with local images, such as Obsidian. You set the folders once during setup. ### How much disk space does it use? A two hour 1080p talk downloads at about 290MB, plus about 50MB of working frames. Only the chosen slides stay in your notes. ### What if the download fails? The most common cause is an outdated yt-dlp. The skill knows the fix and upgrades it before it tries anything else. ### What if my first note is not great? Tell it which slide is missing or which caption is wrong. It re-checks that part of the video and updates the note. ### Does this skill work with Gemini CLI, Codex, Cursor, or GitHub Copilot? Yes, with one caveat. The skill is a SKILL.md folder in the Agent Skills open standard, which Claude Code, Codex CLI, Gemini CLI, Cursor, and GitHub Copilot all load. It is built and tested on Claude Code. Steps that lean on Claude-only features, such as parallel subagents, run one after another or need a manual step in other agents. ### How do I install an agent skill? Copy the skill folder into your agent's skills directory: ~/.claude/skills for Claude Code, or the skills folder your agent documents for Codex CLI, Gemini CLI, Cursor, or Copilot. The agent reads the description and loads the skill when your request matches it. Each skill ships with a setup guide for any tool it needs. [All AI skills](https://atharvashah.com/skill-pack) ยท [Library index](https://atharvashah.com/ai-skills.md)