How to use the Podcast Transcript to Clips Planner
- Paste a transcript with times. Use lines such as 12:30 Speaker: text or 01:02:03 text, an SRT or WebVTT file, or a list of chapter markers. Speaker names and >> change markers are picked up.
- Set the length range. The defaults are 20 to 90 seconds. Raise the minimum for longer reels or lower the maximum for short feeds, then choose how many clips to list.
- Read the ranked windows. Each window shows its start, end, score, the signals that earned it and a suggested caption line. Windows never overlap each other.
- Export for your editor. Download the CSV of start, end and label to mark cuts in your editor, and copy the chapter list into your YouTube description.
How the clip finder works
What it checks
Every possible run of consecutive transcript lines whose total length falls inside your range becomes a candidate. Each candidate is scored on questions (up to 3), numbers (up to 3), emotional words (up to 4), laughter markers (up to 2), speaker changes (up to 3) and a hook-style first line, with a small bonus for lengths near 45 seconds. Mid-sentence starts, filler openings and abrupt endings lose points.
The best candidate is taken first, then the next best that does not overlap it, until your clip count is reached. The caption line is the first question, number-bearing or strongly worded sentence in the window, trimmed to about 110 characters.
Sources and rules
YouTube chapter rules (first timestamp 00:00, at least three timestamps, 10 seconds minimum): YouTube Help article on video chapters, checked October 2026. Vertical 9:16 is the usual shape for short-form feeds; use the linked Video Length and Size Checker for each network's current duration limit instead of relying on a number here.
The emotion word list and the weights are our own heuristics, not a platform rule and not a measurement of what performs.
Limitations
It works from text and timestamps only. It cannot hear tone, music, crosstalk or silence, so a window can start a beat early or late. Auto-generated transcripts with wrong words lower the quality of every signal. Always scrub the audio around each start and end before you export a clip.
Frequently asked questions
How does it decide which moments are clip-worthy?
It slides a window across the transcript and scores each one with simple rules: questions, numbers and percentages, a list of strong emotional words, laughter markers such as [laughter], speaker changes and a hook-style opening line. It lowers the score when a clip starts mid-sentence or ends without closing punctuation. It does not listen to the audio or judge whether a moment is actually good, so treat the ranking as a shortlist.
Which timestamp formats work?
hh:mm:ss, mm:ss, bracketed stamps like [12:30], SRT, WebVTT with or without voice tags, and plain chapter markers. When a line has no end time, it ends where the next line begins, and the final line gets 5 seconds.
Why are my clips shorter or longer than I set?
Clip boundaries snap to transcript lines, so a window is always whole lines. If your transcript has very long lines, a window may be forced near the limit. A single line longer than the maximum is trimmed to its opening and marked as a long source line, since the tool cannot tell where inside it to cut.
What does the YouTube chapter list follow?
YouTube asks for the first timestamp to be 00:00, at least three timestamps in ascending order and at least 10 seconds per chapter, per YouTube Help. If your input looks like chapter markers they are used as given. Otherwise chapters come from the start of each suggested clip, with Intro added at 00:00. Always edit the labels.
Is the CSV a real EDL file?
No. It is EDL-style: start and end timecodes with a label and duration, readable in a spreadsheet or by most markers import features. A true CMX 3600 EDL needs a frame rate and reel names, which a transcript does not contain.
Is my transcript uploaded?
No. Parsing and scoring run in your browser and nothing is sent to a server, so unreleased episodes stay on your device.