TikTok transcript

Caption track first, speech recognition second

If TikTok already stores a caption track on the video, from the creator or TikTok's own automatic captions, we read that file. If there is none, we download the video, keep only the audio track, and send it to speech recognition on Workers AI, the platform this site runs on. The timings come from the model's own word list. That second route is the point of this page: a TikTok video with no caption track still returns its spoken words as text, because the words are written from the audio, not read from anything the creator typed.

Measured on 18 September 2026: a 32-second video arrived as an 8,507,135-byte MP4; the audio-only file built from it was 132,707 bytes, and the recognized text matched the video's own caption track word for word.

Long audio goes to the model in pieces of about 45 seconds. Speech recognition handles up to 20 minutes per video; a caption track has no length limit here.

What a 20-link test returned

On 18 September 2026, 11 of 20 links came back with text: 4 from an existing caption track, 7 from speech recognition.

Nine did not: seven were videos TikTok would not serve at all (three of those were IDs we made up on purpose to test that path), one had audio that could not be downloaded, and one hit a failed recognition call. That is a 20-link sample, not a success rate.

Music and speech recognition

When a clip is mostly music, the model may return a repeating line or a stock phrase instead of real words. We saw that on a music-heavy clip in the same run. Clips that are mostly talking work best.

When there is no text

• The video would not load: private, removed, or not a video link.
• No caption track, and the video file could not be downloaded at that moment. Worth trying again.
• No caption track, and the audio had no recognizable speech: a music-only or silent clip.
• No caption track, and the recognition call failed. Worth trying again.
• No caption track, and the audio is longer than 20 minutes.
• Short links from the app (vt.tiktok.com, vm.tiktok.com and tiktok.com/t/) are followed to the video automatically. Photo posts and profile links are not videos, and the page says so instead of producing filler text.

TXT, SRT or VTT

Whatever you call the job — TikTok transcripts to keep, one TikTok video transcript for a quote, or TikTok to text — the result ends in the same three files. TXT is the readable script with no clock. SRT is the subtitle file an editor expects, with numbered cues and times like 00:00:01,280. VTT holds the same cues in WebVTT form for web players, with times like 00:00:01.280.

Paste a link and take a file with you on the tiktok transcript generator.

When the job is a clock-free wording you can set beside your own draft, use the tiktok script extractor.