# Changelog of TikTok Transcript Scraper — TikTok Video to Text (`lergassy/tiktok-transcript-scraper`) Actor

- **URL**: https://apify.com/lergassy/tiktok-transcript-scraper/changelog.md
- **Full Actor documentation**: https://apify.com/lergassy/tiktok-transcript-scraper.md

## Changelog

### 2026-10-13

- A bare TikTok username in **TikTok profiles** (for example `duolingo`, without `@`) now works like `@duolingo`; before, it returned an "is not a valid URL" error row.
- Every run now ends with a status message saying how many videos were transcribed and how many inputs failed, including runs where nothing was found ("Nothing was charged").

### 2026-10-09

- On-screen text is read about twice as fast per frame, and runs that read on-screen text get a full CPU core (4 GB by default), so two clips with on-screen text no longer take ten minutes.
- Music-only clips whose speech recognition returned a stock line in Korean, Japanese, Chinese, German, French or Spanish ("please subscribe" and similar) are now marked as no speech instead of being returned as a transcript.
- On-screen text is read one video at a time with the container's own CPU share, and never past the run's time limit: a video whose on-screen text cannot be read in time still returns its transcript (with `onScreenTextError`, on-screen text not charged) instead of the run timing out.

### 2026-10-02

- Default memory now scales with your input (number of videos, on-screen text on or off), so small runs use less memory.

### 2026-09-27

- Rows are delivered as soon as they are ready instead of at the end of the run; videos from one profile arrive together once the profile is done.
- Runs stop cleanly at your maximum cost per run instead of being aborted; nothing beyond your limit is written or charged.

### 2026-09-26

- Captions first is the new default: when TikTok has its own subtitles for a video they become the transcript, which is faster and keeps every repeated line; videos without them go to speech recognition.
- New "Find viral videos" option (on by default, free): adds author, views, likes, comments, caption and publish date, and flags videos that did 3x or more of the author's usual (`vsAuthorMedian`, `isOutlier`).
- Private or removed videos return a free error row right away; music-only clips no longer get invented text like "Thank you." and are not charged for audio.

### 2026-09-24

- Repeated lines in TikTok and Instagram captions are kept instead of being merged away.

### 2026-09-23

- Runs start much faster thanks to a smaller container image.

### 2026-09-21

- Initial release: turns public TikTok videos (and Instagram Reels or YouTube Shorts links) into text: transcript with timestamps, optional on-screen text and the first-3-second hook, from video links or a handle like @duolingo; no login or cookies needed.