One of the most time-consuming parts of repurposing long-form content is figuring out which sections are worth cutting into shorts. You have a 30-minute podcast episode or a 20-minute tutorial, and somewhere inside it are three or four moments that would work as standalone clips.
The problem is finding them.
What Makes a Moment Worth Clipping
Not every interesting segment makes a good short. The moments that perform well on platforms like YouTube Shorts, TikTok, and Instagram Reels tend to share specific characteristics:
- They are self-contained. A viewer who has never seen the full video should be able to understand the clip without additional context.
- They have a clear emotional arc. The best clips contain a setup and a payoff within 30–60 seconds. A surprising stat, a strong opinion, a funny reaction.
- They start strong. The first two seconds need to justify why someone should stop scrolling. If the opening is a slow transition or an “um,” the algorithm buries it.
When you are scrubbing through raw footage manually, you tend to pick moments based on what you found interesting during the edit. But what you find interesting as the editor and what performs well as a standalone clip are often two different things.
The Manual Approach
The traditional method looks like this: watch the full video, take notes on timestamps that feel promising, go back to each one, evaluate whether it works out of context, trim the in/out points, and export. For a 30-minute video, this process can take an hour or more.
Some editors speed this up by working from the transcript instead of the video. Reading text is faster than watching footage, and you can scan for strong quotes, surprising data points, or emotional shifts without sitting through the entire runtime.
Automating the Search
This is exactly what the Find Shorts tool inside SmoothyEdit is built for.
You provide your video content — either by pasting a YouTube link, uploading a subtitle file, or dropping in audio for transcription. The AI reads through the entire transcript and flags the segments that meet the criteria for strong short-form content: self-contained narratives, emotional hooks, and clear payoffs.
For each result, you get the exact start and end timestamps, along with a brief explanation describing why that particular section was selected. From there, you can copy the timestamps or download the full list.
Sending Markers Directly to Premiere Pro
If you use Adobe Premiere Pro, there is a faster path than copying timestamps and manually scrubbing to find them on your timeline.
Studio — the optional cloud add-on — connects the webapp to the SmoothyEdit desktop app and on to Premiere Pro. After running the Find Shorts tool, each result includes a Send to Premiere button. Clicking it drops color-coded markers at the exact start and end points on your open sequence. (Find Shorts uses cloud AI and sends the relevant transcript or source input; audio upload depends on the input you choose and your plan; the app’s multicam, silence, caption and compression tools stay local.)
Instead of watching the footage to find the cuts, you jump from marker to marker using Premiere’s standard marker navigation, make your slices, and start editing the clips. The Best Shorts Markers tool (part of Studio) also lets you run the detection from within the SmoothyEdit app without switching to the browser.
You can download the app from the Premiere Pro plugin page.
Getting Better Results
Two things that consistently improve the quality of the AI output:
- Use accurate transcripts. If you are working from auto-generated YouTube captions, the AI may misinterpret sections where the captions are garbled. Uploading a subtitle file or a clean audio file for transcription produces significantly better results.
- Longer content gives more options. A 5-minute video might only yield one or two viable clips. A 30-minute video might yield eight to ten. The more material the AI has to work with, the more selective it can be.
The goal is not to automate your editorial judgment. The goal is to remove the tedious step of watching an entire video just to find four timestamps. Once the tool surfaces the candidates, you still decide which ones actually make the cut.