Turn videos into everything else

The repurposing primitives

• transcript — full speech-to-text with timestamped segments. Complete coverage on TikTok and Instagram; caption-dependent on Facebook and X (the endpoint pages say exactly what to expect).
• summary — a short AI summary when you need the gist, not the words.
• channel-videos / channel-reels — enumerate a channel so you can batch-process the back catalog.

Video to blog draft, end to end

import httpx

BASE = "https://api.scrapersocial.com/v1"
H = {"Authorization": "Bearer sk_live_..."}

t = httpx.get(f"{BASE}/tiktok/transcript",
              params={"url": video_url}, headers=H).json()

draft = llm.complete(
    f"Rewrite this video transcript as a 600-word blog post. "
    f"Keep the speaker's voice:\n\n{t['data']['text']}"
)

Timestamped segments in the same response let you generate chapter markers, pull quotes with timecodes, or subtitle files without a second call.

What a real pipeline costs

The math is simple enough to do in your head. A TikTok transcript is 8 credits, so turning one video into a blog draft costs about 4 cents before your LLM bill. A creator's 100-video back catalog is 800 credits, or $4. Instagram runs richer: 40 credits per reel transcript, so the same 100-reel catalog is $20. If you only need the idea rather than the words, summaries cost a couple of credits more than transcripts and save you the LLM step entirely.

Batch a whole channel

List a channel's videos, then transcribe each one. The transcribe-a-channel recipe shows the full loop with pagination, and async jobs with webhooks handle catalogs too big for a synchronous loop. Transcripts are cached forever, so re-running a pipeline over the same videos returns in milliseconds instead of waiting on speech-to-text again.

Where the output goes

Teams wire this into three places, usually in this order: a CMS draft queue (video in, draft post out), a newsletter tool via Zapier or Make (weekly digest of what a set of creators said), and clip selection (scan segments for quotable moments, cut around the timecodes). The TikTok and Instagram transcript endpoint pages have the request shapes for all of it.