Content repurposing with transcripts: one video, ten posts

November 8, 2026

Repurposing works when you cut in the right place. A 45-minute episode usually holds five to ten moments that make sense on their own, and the transcript tells you where they are faster than scrubbing a timeline. Pull the transcript with timestamps, score short windows, cut the winners, then check which clips earned their place.

Who does this

• Podcasters and YouTubers turning each long upload into Shorts, reels and TikToks.
• B2B marketing teams turning webinars and LinkedIn videos into posts and threads.
• Agencies running the same pipeline for many clients every week.
• Newsletter writers turning a video into a written summary with quotes.

The data, step by step

Two limits to plan around. The Instagram transcript endpoint handles reels under 2 minutes. YouTube Shorts can be up to 3 minutes long, which gives you room for a complete answer rather than a teaser.

Which long videos to repurpose first

Start from performance, not from the newest upload.

• Repurpose index = video views / median views of your last 20 uploads. Anything above 1.5 has proven topic demand.
• Age filter. Skip videos under 7 days old; their view count is still moving.
• Evergreen check. Drop videos tied to a date or a news event unless the clip still makes sense now.

The clip-finding method

1. Fetch the transcript. Each segment has text, start_ms and end_ms.
2. Slide a 45-second window across it in 15-second steps.
3. Score each window: +2 if it contains a number, +2 if it answers a question ("how", "why", "the reason"), +1 for a strong opener ("the mistake", "most people"), −3 if it points back ("as I said", "earlier").
4. Keep the top five windows that do not overlap. Snap the start to the nearest sentence start.
5. Hand the timestamps to your editor, or to a cutting tool, with the window text as the draft caption.

import os, re, requests

KEY = os.environ["SCRAPEIO_API_KEY"]
r = requests.get("https://scrapeio.com/v1/youtube/video/transcript",
                 params={"url": "https://www.youtube.com/watch?v=VIDEO_ID", "language": "en"},
                 headers={"x-api-key": KEY}, timeout=60).json()
segs = r["data"]["segments"]

def score(text):
    t = text.lower()
    s = 2 * bool(re.search(r"\d", t))
    s += 2 * any(w in t for w in ("how ", "why ", "the reason"))
    s += any(w in t for w in ("the mistake", "most people", "here's"))
    s -= 3 * any(w in t for w in ("as i said", "earlier", "like i mentioned"))
    return s

WIN, STEP = 45_000, 15_000
end = segs[-1]["end_ms"]
windows = []
for start in range(0, end, STEP):
    text = " ".join(s["text"] for s in segs if start <= s["start_ms"] < start + WIN)
    windows.append((score(text), start, text))

picked = []
for sc, start, text in sorted(windows, reverse=True):
    if all(abs(start - p[1]) >= WIN for p in picked):
        picked.append((sc, start, text))
    if len(picked) == 5:
        break
for sc, start, text in sorted(picked, key=lambda p: p[1]):
    print(f"{start // 60000}:{start // 1000 % 60:02d}  score {sc}  {text[:90]}")

The same windows feed the written formats: one window becomes a LinkedIn post, three become an X thread, and the full transcript becomes a newsletter summary with direct quotes.

Worked example: cost per month

A channel posts four long videos a month and clips them to TikTok, Instagram, YouTube Shorts and LinkedIn.

36 credits is about $0.07 on the $47 pack. An agency running this for 12 clients spends 432 credits a month, about $0.81. The data is not the bottleneck. Editing hours are, which is why cutting in the right place first time is where the method pays.

Measuring the clips

• Plays per follower per platform, compared only with your own earlier clips on that platform.
• Clip hit rate = clips above your platform median / clips posted. Track it by window score to check that the scoring rule predicts anything.
• Click-back rate, if you link the full video: long-video views in the 48 hours after a clip, compared with the week before.

Pitfalls

• Auto-captions mangle names. Fix product and person names before the text becomes a caption.
• No captions. Some videos have none. That call fails and is not charged; fall back to speech-to-text on your own file. Captions versus speech-to-text covers the trade-off.
• Rights. Repurpose content you own or have permission for. Competitor transcripts are for studying structure, not for copying lines.
• Edited re-uploads. If you trimmed the video after upload, timestamps from the original file will not match.

Run it from an AI agent

claude mcp add --transport http scrapeio https://scrapeio.com/mcp --header "x-api-key: $SCRAPEIO_API_KEY"

Then ask: "Get the transcript of this video with youtube_transcript. Find five 45-second segments that stand alone, with start times. For each, write a TikTok caption, a LinkedIn post of under 120 words and a hook line. Quote the transcript exactly." The agent does the scoring in plain language, and you keep the timestamps for the edit. More on transcripts: the YouTube transcript guide and studying reel hooks.

How to do it with ScrapeIO

One key gives you transcripts for YouTube, TikTok, Instagram, Facebook, LinkedIn and X, plus the post lists to measure what you publish. Most calls cost 1 credit, the Instagram transcript costs 2, and credits never expire, so a small monthly workflow like this one runs for years on one pack.

Sources

Checked on 8 November 2026 from each vendor's own pages. Prices and limits change; follow the links before you decide.

• https://support.google.com/youtube/answer/10059070