InsightSocial API
How-to

Read on-screen text from TikTok videos

Get the text burned into a TikTok video, read by OCR, from the video URL. 50 credits per video.

To read the text shown on a TikTok video, call GET /v1/tiktok/video/screen-text with the video's url: it returns TikTok's native text stickers plus AI OCR of the cover frame, so captions burned in with an editor are caught too. It costs 50 credits per video.

Cost

50 credits per video. Reading 100 videos costs 5,000 credits. Asking for the same URL again within 6 hours is answered from the shared cache for 2 credits. Failed calls and empty results are free.

Which text does each endpoint read?

A post carries text in three places, and each has its own endpoint.

TextEndpoint
CaptionGET /v1/tiktok/post (10 credits)
Spoken wordsGET /v1/tiktok/post/transcript (100 credits)
Text on screenGET /v1/tiktok/video/screen-text (50 credits)

On TikTok, text typed with TikTok's own text tool is also on GET /v1/tiktok/post as post.ext.on_screen_texts. Use screen-text when you also need text burned into the video by an outside editor, which only OCR can see.

How do I find the URL?

Copy the video's link from the app or browser: a /video/ or /photo/ link, such as https://www.tiktok.com/@stoolpresidente/video/7499229683859426602.

Make the request

cURL
# TikTok video
curl "https://api.insightsocial.app/v1/tiktok/video/screen-text?url=https://www.tiktok.com/@stoolpresidente/video/7499229683859426602" \
  -H "x-api-key: $INSIGHTSOCIAL_API_KEY"

What comes back

A real response from 2026-10-14, complete:

TikTok response
{
  "success": true,
  "endpoint": "/v1/tiktok/video/screen-text",
  "schema_version": "2",
  "data": {
    "video_id": "7499229683859426602",
    "url": "https://www.tiktok.com/@stoolpresidente/video/7499229683859426602",
    "texts": ["Delicious Pizza"],
    "native_texts": null,
    "ocr_texts": ["Delicious Pizza"],
    "source": "ocr",
    "media_type": "video",
    "frame": "cover"
  },
  "unavailable": [],
  "credits_used": 50
}
FieldWhat it holds
data.textsAll the text found on the video, native and OCR combined.
data.native_textsText added with TikTok's text tool. null here, so everything came from OCR.
data.ocr_textsText read by AI from the image.
data.sourceWhere the text came from: ocr in this response.
data.frameWhich frame OCR read: cover.

What are the limits?

  • The cover frame only. OCR reads only the cover frame of a video, so burned-in text that appears later in the video is not read.
  • Lines, not sentences. OCR returns text in the lines it sees, so one sentence can arrive as two entries, and an entry can contain a line break ("Happy\nBirth Day"). Join them yourself if you need full sentences.
  • Not the caption, not the audio. Use the post and transcript endpoints in the table above for those.

What can I build with it?

  • Hook analysis. Collect the text on the cover frame of top-performing videos in a niche and compare which hooks get the most views.
  • Searchable archive. Store caption, transcript and on-screen text together so a search finds a post by any words it shows.

On this page