TikTok Transcript Scraper — TikTok Subtitles to Text, No Login
Pricing
from $1.00 / 1,000 videos
TikTok subtitles and transcripts for a creator's videos: type a handle and get every video as a row — views, likes, comments, shares, saves, caption, hashtags, duration and a playable MP4 link — with an opt-in transcript from TikTok's own captions or speech-to-text. No login.
All notable changes to this actor are documented here. Each entry matches one build on the Builds tab.
A media link that does not come from TikTok's own servers is now refused before it is fetched, instead of being downloaded, transcribed and charged. Every link this actor opens for a transcript is written by the data feed it buys from, not by you: the video's playback renditions, TikTok's own caption file, the video page address handed to the fallback resolver, and the address that resolver prints back. Nothing checked where any of them pointed, so a wrong or tampered answer would have been downloaded from whatever host it named, sent to the speech service and billed to you as a transcript. All four are now checked against TikTok's own servers first, and a link that is not one is never opened: the video row is still delivered, the transcript is uncharged, and the row says the link was refused and names the host — it does not say the video has no audio, because those are two different things. The checks are independent, so a refused playback link on a real TikTok video page still gets the fallback resolver, and a refused caption file still falls through to speech: nothing that used to produce a transcript stops producing one. No price, input field, output column or charged event changed.
A run that collected nothing now tells you what to change, instead of only that it collected nothing. A handle this actor could not read ended the run cleanly, so the run page said "Delivered 0 videos of 1 asked for across 0 profiles." and stopped there — 52 characters, no cause and no next step. Two things were wrong. The count of values it could not read was kept on the run's own record and never printed, so a run whose only fault was a mistyped handle named nothing at all. And nothing on the page said what to put there instead. The line now carries that count, in words rather than internal codes, and on a run that delivered nothing it adds one plain sentence for the biggest cause: the box to use and a working handle to copy, what a private or deactivated account means, what a creator TikTok lists no videos for means, and which of those a second run would change. An order the run could not use gets its own sentence naming the two that work, rather than advice about handles that were never the problem. A feed read that did not go through is still never reported as "this creator has no videos" — it does not know. A run that did deliver is unchanged. No price, input field, output column or charged event changed.
The page now opens with the name you clicked, the price, and a real video row. The first line of the page called this actor something other than its store title, what a run costs was four screens down, and the page listed every column without ever showing one. The heading is now the store title, the opening paragraph carries the per-1,000 price on the cheapest paid tier and on the Apify free plan, and a real delivered row — the caption, hashtags, post date, duration, plays, likes, comments, shares, saves and the sound — sits above the column table, built by this actor's own parser from a harvested profile feed. No price, input field, output column or charged event changed.
profile, username, usernames,
handle, handles, creator, creators, profileUrl or profileUrls, or a profile link as
url, urls or startUrl, and the run reads them instead of collecting nothing: the values are
used exactly as sent, a creator you name in both places is collected once, and one uncharged note
names the field to use next time. Nothing about what is charged changes.A brief "slow down" from the data feed no longer costs you the rest of the creator. When the feed behind this actor answered with a rate limit, a server error or an unreadable reply part-way through a creator's videos, the run used to give up on that creator within a few seconds: you kept the videos collected up to that point and the rest of your result limit came back as something to re-run. From this build the run waits the refusal out, up to twice per creator, and asks for the same page again; a feed that clears carries on to your result limit, and only a feed that still refuses afterwards ends the walk. The waiting is bounded by your run clock — no wait is taken that the remaining time cannot fit — and it stops immediately at your maximum cost per run. A verdict about the account itself (not found, private, no videos) is never waited on, because waiting cannot change it.
A creator that really was cut short is unchanged: the videos already collected are still delivered, and nothing that did not arrive carries a result fee.
A limit of ours is never published as a verdict about the creator. A walk that ends at this actor's own per-creator page limit now says exactly that — how far it read, that the feed was still listing more, and that it is a limit on this run rather than an answer about the account. The platform-verdict wording is kept for the one ending that earns it: TikTok's own feed closing. And where the result limit went unfilled because the videos on those pages were ones your account already has, the row names the switch that would have filled it.
The run record names the same cause the row does. A walk this actor's own per-creator page limit cut is recorded as its own clean stop rather than as "no videos" — the wording a buyer reads, the run page's count and the machine-readable outcome record now agree about which limit ended the walk, so a run you read through the API says the same thing as the row in front of you. Nothing about what is charged changes: those videos were never collected and are never billed.
Your rows carry the data and nothing about the fetch behind it. An address fragment that could reach a row is now stripped in every shape.
First release.
transcript, transcriptLanguage, timestamped transcriptSegments and transcriptSource — which says whether the words came from TikTok's own captions or from speech-to-text.maximum on any number and no minimum above 1, so a caller asking for 20,000 videos per creator — or a 5-second run clock — still starts a run: the actor's own ceiling binds (2,000 per creator, 20,000 rows, a 30-3,600 second clock) and one uncharged row says what was asked and what bound it. A schema ceiling is validated by the platform before a container boots, so it returns an error and no run at all — invisible to the buyer and to us.sortBy: "latest" walks the profile grid newest first; sortBy: "popular" walks TikTok's own most-played ordering for that creator, which is a different set of videos.video, from $1.00 per 1,000 videos, platform usage included; transcript, only on a video whose spoken words were actually produced; long-video-surcharge, one per started minute beyond the three minutes included in the transcript price. No start fee, and no second charge for "details".