How to Repurpose One Long Video Into a Week of Short Clips With AI

Get a timestamped transcript of your long video, then ask any AI assistant to find the 5-8 best standalone moments and return a start and end timestamp, a hook, and a caption for each. Cut those exact ranges into vertical clips, eyeball them, and schedule one a day. One 40-minute recording usually holds a full week of shorts, and the AI does the tedious part: hunting for the moments worth posting.

You already make long content: a podcast, a webinar, a client training, a talking-head explainer. Inside that one 40-minute file are several complete thoughts that each stand on their own. Done by hand, finding them means scrubbing the timeline over and over, which is why most people record long and post nothing. AI changes the math: it reads the whole transcript in seconds, points you to the moments worth posting, and drafts a hook for each. You stay the editor. Strip away the apps and the loop is three moves you reuse on every video:

• Transcribe — turn the video into text with timestamps.
• Find — ask AI for the 5-8 best standalone moments, each with a start and end time, a hook, and a caption.
• Cut — clip those exact ranges into vertical video and review each one.

Learn it once and you stop wondering what to post. The same loop fits a podcast, a webinar, or a webcam explainer; the tools can change but the three moves do not.

You need the words and the times they were said. If your video lives on YouTube, the auto-transcript already includes timestamps you can copy. Most recording and meeting tools export a transcript too, and free transcription sites will turn an uploaded file into a TXT or SRT with times. The format that helps most is one with timestamps every line or two, because those times are what your AI will hand back as cut points. Skip the urge to paste a raw, time-free wall of text. Without timestamps the AI can find a good line but cannot tell you where it lives in the video, and you are back to scrubbing.

Paste the transcript into ChatGPT or Claude and ask for self-contained moments, not topics. One line that works:

From this timestamped transcript, find the 6 best standalone moments for short video. For each: start and end timestamp (30-60s), a one-line hook, and a 2-line caption.

The words "standalone" and "30-60s" do the heavy lifting. They push the AI toward clips that make sense with zero context, which is the only kind that performs cold on a feed. Ask it to favor a strong opinion, a surprising number, a clear how-to step, or a short story. If a moment needs the previous ten minutes to make sense, it is not a clip, and a good prompt tells the AI to skip those.

The AI hands you a tidy list. Do not trust it blindly. Read each suggested moment in the transcript and ask one question: would this make sense to someone who has never heard of me? If a clip opens mid-argument or leans on a setup that got cut, drop it or widen the time range to include the setup. Tighten greedy ranges too. A moment marked 30 seconds often has a slow five-second wind-up you can trim so the point lands faster. This is the step beginners skip, and it is the difference between clips that feel deliberate and clips that feel like a robot chopped your video at random.

Now turn each approved range into vertical video. Open your editor, set the in and out points to the timestamps you confirmed, crop to 9:16, and burn in captions so the clip works on mute. Most short-form views happen with the sound off, so on-screen text is not optional. Do this once per moment and you have your week. If hand-cutting six clips feels slow, an all-in-one tool like OpusClip can take the long video, find moments, and auto-produce captioned vertical clips in one pass, which you then trim. Manual teaches you the craft; automation saves time once you know what good looks like.

The first second decides whether anyone watches, so the on-screen hook matters more than the edit. Use the AI's draft hook as a starting point, then sharpen it into a real promise or an open question: "The pricing mistake that cost me a client," not "Here are some thoughts on pricing." Pair each clip with the two-line caption the AI wrote, add two or three relevant tags, and schedule one clip per day. Spacing them out keeps your feed alive all week from a single recording, and gives each clip its own shot instead of burying five in one afternoon.

A few rules keep this from turning into spam. Post the strong moments and bin the weak ones, even if it leaves you with four clips instead of six; quantity with no substance trains people to scroll past you. Never let AI invent a quote or a stat that was not in the video, your clips are your own words, so they should be exactly that. Watch every clip start to finish before it goes out, because a bad cut or a wrong caption does more damage than a missed post. And keep one prompt and one editor you reuse every time, so the loop gets faster with each video instead of starting from zero.

Take one long video you already have. Grab its timestamped transcript, paste it into ChatGPT or Claude, and run the "find 6 standalone moments" prompt above. Read the list, keep the three that would make sense to a stranger, and cut just those into vertical clips with captions. You will have three ready-to-post shorts from one recording in well under an hour, and the exact loop you can run on every video from here on.

Get the transcript of your most recent long video today — that is Step 1 and it costs nothing. Run the clip-finding prompt on it and cut one short. One good clip proves the pipeline; the week of posts follows.