Get transcripts from YouTube, TikTok and Instagram with one API

November 7, 2026

Most of what a video says never makes it into its caption. Transcripts turn video into text you can search, summarize, and feed to a model. Each platform hides them differently. ScrapeIO gives you one call per platform and the same response shape.

The endpoints

Each one takes the post's url. YouTube also takes language, for example en.

curl "https://scrapeio.com/v1/youtube/video/transcript?url=https://www.youtube.com/watch?v=dQw4w9WgXcQ&language=en" \
  -H "x-api-key: $SCRAPEIO_API_KEY"

Good to know

• YouTube and TikTok transcripts come from the video's captions, including auto-generated ones. Videos with captions turned off have none, and the call is free.
• Instagram transcripts work on reels under 2 minutes.
• Not found and private answers cost 0, so a batch over a mixed list only charges for real transcripts.

From a channel to a searchable library

Combine a list call with the transcript call. In Python:

import os, requests
H = {"x-api-key": os.environ["SCRAPEIO_API_KEY"]}
API = "https://scrapeio.com"

videos = requests.get(f"{API}/v1/youtube/channel-videos", params={"handle": "mkbhd"}, headers=H).json()["data"]["items"]
for v in videos[:10]:
    t = requests.get(f"{API}/v1/youtube/video/transcript", params={"url": v["url"]}, headers=H).json()
    if t.get("success"):
        print(v["caption"], len(str(t["data"])))

Ten videos cost 11 credits: one for the list, one per transcript.

With an agent

Connect the MCP server and ask: "Get transcripts of the last 10 videos on this channel and tell me which products come up more than once." The agent calls youtube_channel_videos and youtube_transcript itself and quotes the transcript when it answers.