TikTok Video Transcript — AI Video to Text & SRT
Pricing
from $7.20 / 1,000 audio minutes
Get a TikTok transcript from any public video with AI speech-to-text, even without captions. Paste video or share links; get timestamped text, SRT/VTT subtitles for vertical video, speaker labels and RAG-ready JSONL. No login or API key. $0.013 per 1-minute video + $0.005 per run.
Public TikTok video links, one transcript per URL. Accepted forms: tiktok.com/@user/video/, vm.tiktok.com/, vt.tiktok.com/ and tiktok.com/t/. Profile URLs go in TikTok profiles below; photo, tag, sound and live URLs are rejected. Submit up to 1,000 URLs, or leave this list empty and fill in TikTok profiles or Dataset ID below.
Optional BCP-47 language code, for example en, ru or pt-BR, that guides recognition. Omit it for automatic language detection; an empty string is invalid. Detection and language support depend on the speech engine.
Choose no timestamps, segment timestamps (subtitle-style lines) or word timestamps. Segment and word output includes SRT and VTT when valid cues are available. Subtitle profiles and RAG passages need segment or word timing. Timed lines are in the segments array (word timing in words): read the whole field (fields=segments), because dotted paths such as segments.t0 return nothing.
Caption layout and an advisory QA report. Short video (28 characters per line, 0.5–4 s cues) suits vertical clips and is the default whenever timestamps are on and this field is empty. Requires segment or word timestamps: clear it when you choose no timestamps. Cues over the line or duration limits are split inside their segment's time range; segments keep their original times, and warnings do not stop transcription.
Identify speakers (S1, S2, …). Billed at the premium per-minute rate. Whole-video labels are S1/S2; independently processed chunks use C1_S1/C2_S1, so read speaker_scope in each result.
Return rag_chunks and transcript-NNN.rag.jsonl with passage text, original-video seconds, source URL and versioned document identity. Requires segment or word timestamps. Embeddings and indexing remain downstream steps.
Public profile URLs such as https://www.tiktok.com/@bbcnews. The newest videos are listed (pinned videos are sorted by date), and each is processed exactly like a video URL above; each row names the profile it came from. TikTok profile listings fail intermittently, so each is tried up to 3 times; a profile that still fails becomes an error row that is not charged. Up to 100 profiles. The example URL above is processed and charged too: remove it if you want only listed videos.
List at most this many of the newest videos from each source (up to 200). Each listed item is processed and billed like a URL you submit; listing itself is free. A run holds at most 1,000 items in total, counting URLs and dataset rows.
Optional date as YYYY-MM-DD (UTC). Items published before this day are left out. TikTok listings give exact dates.
The example URL above is processed and charged too: remove it if you want only the dataset's items. Pick the dataset, or paste its ID, from a TikTok Scraper run (for example clockworks/tiktok-scraper) to transcribe every video it found. The Actor reads the rows with your run's own token and processes each row's URL like a URL above, after the listed URLs. Rows without a usable URL become uncharged error items. A named dataset works as username~dataset-name.
Optional. The row field that holds the URL, for example webVideoUrl, or a dotted path such as video.url. Leave it empty to use the first accepted URL from webVideoUrl, url, postUrl, videoUrl, inputUrl or link.
Read at most this many dataset rows, in dataset order. A run processes at most 1,000 URLs in total, counting the list above.
Remember every item that was delivered and charged, in a key-value store in your Apify account, and skip those items in later runs. Use it with a schedule to get only new videos or episodes. Items that failed are tried again next time. When every item was already processed, the run ends without any charge. If the store cannot be read, the run stops before any charge and the next run tries again.
Optional. Keeps a separate memory for this schedule, for example my-weekly-podcasts. The store is named polyglot-processed-tiktok--, and other Actors never share it; without a name every run of this Actor shares polyglot-processed-tiktok. Delete the store in Apify Console (Storage, Key-value stores) to start over. Used only with the option above.
