You record, you post once, you move on. The transcript holds 5 more pieces of content: chapters, themes, quotes, a quote card, and a thumbnail. Claude pulls all 5 without you re-watching a minute.
Claude reads text, and for a recorded video the transcript already holds everything worth pulling. Your video app makes one automatically: on YouTube it's the auto captions, on most podcast and live platforms there's a transcript download in the edit menu. What you want is the timestamped version, the file where every line has a time next to it.
Download it, open a Claude chat, and run the 3 prompts below in order, in the same chat. Each one builds on the last.
Paste 2 or 3 lines of what YOU thought mattered in the recording: the moment that surprised you, the point you want highlighted. The assets come out sounding like your read of the conversation instead of a generic summary.
15 to 20 chapter markers you paste straight into YouTube, Apple Podcasts, or Spotify. The no-rounding line is the fix that makes them land on the right moments.
Why it works. Left alone, AI estimates timestamps and your chapters drift. Telling it to use the exact start times from the transcript, with no estimating, keeps every marker honest.
Here is the timestamped transcript of my video: [paste transcript]. Find the 15 to 20 key moments. For each one give me the timestamp in MM:SS, using the exact start time from the transcript segment. Do not estimate or round any timestamp. Add a short description of what's happening at that point, and format the whole thing as a clean list I can paste into YouTube, Apple Podcasts, and Spotify as chapters.
The big themes become next week's post ideas. The word-for-word quotes become your pull quotes, and the strongest one becomes the quote card.
Why it works. Word for word is the rule that matters. A paraphrased quote is a caption; a real quote in real words is an asset you can put a name under.
From the same transcript, pull the 3 or 4 biggest themes of the conversation. For each theme give me a one-line headline and the single strongest quote that captures it, word for word from the transcript, under 30 words. Then tell me which one quote is strong enough to stand alone as a quote card.
Claude can't draw your thumbnail, and it doesn't have to. It writes the brief, and any image tool renders it.
Why it works. 4 words or less is a real thumbnail rule: at feed size, long headlines vanish. The brief plus a short headline hands the design job to whichever tool you like.
Based on this video's strongest moment, give me a thumbnail concept: what's in the frame, the emotion on the face if there is one, the colors, and a headline of 4 words or less. Keep it simple enough to read at phone size. Then write the concept as a one-paragraph image prompt I can paste into an image tool.
Chapters, themes, pull quotes, the quote card, and the thumbnail. One recording, one sitting, zero re-watching. The themes alone usually carry your posting for the week.
Once the 3 prompts feel routine, hand the routine to Claude for good. This builds the whole workflow into a skill: transcript in, all 5 assets out, in your brand every time.
Create a skill called video-assets. Whenever I give you a timestamped transcript, produce all of this in one pass: 15 to 20 chapter markers using exact transcript start times with no estimating or rounding, the 3 or 4 biggest themes with headlines, the strongest quotes word for word under 30 words each, your pick for the one quote-card quote, and a thumbnail concept with a headline of 4 words or less. My show or brand: [your name, your topic, your colors or style notes]. Format everything ready to paste. Walk me through saving the skill to my account, step by step, beginner level.
The bracket carries your identity: your name, what your videos are about, the look you want thumbnails to lean toward. The more real detail you give it once, the less every future transcript needs from you. To change anything later, just tell Claude what to adjust.