The on-screen text of one TikTok video, one item per line, in reading order.

The on-screen text of one public TikTok video: the words burned on top of the frames, read logged-out from the video page itself. Takes the FULL video address -- https://www.tiktok.com/@handle/video/1234567890123456789 -- and not a bare id, because the page needs the handle in the path. Returns one item per line, in the order the document declares them, which is reading order on screen. This is NOT the caption and NOT the spoken audio: the caption is the text field of tiktok/post, and the spoken audio is a separate WEBVTT asset served by tiktok/post_transcript. Three different texts live in the same document and mixing them up would return plausible wrong data. Measured on 2026-10-01 across 10 reachable videos: 7 carry at least one line of on-screen text and 3 carry none. A video that exists and has no on-screen text answers with items: [] and IS charged -- the question was answered and the page was downloaded. A restricted, missing or unclassifiable video answers 404 and costs nothing. metrics, published_at and thumbnail_url are null on every item: a line of text has no date, no picture and no counter, and returning zeros would claim it has them. Costs 1 credit and caches for an hour. The cache is per endpoint and params, so asking for post and screen-text on the same video downloads the page twice.