What is a TikTok transcript API?
A TikTok transcript API takes a public TikTok video URL and returns the spoken or captioned text as structured data. Social Fetch exposes GET /v1/tiktok/videos/transcript: WebVTT in data.transcript when a caption track exists, with an optional useAiFallback path for speech-to-text when captions are missing.
What the response contains
Branch on data.lookupStatus before you write rows. found means you got text; not_found means the lookup finished and there was nothing to return. Photo carousel URLs are not videos and fail with transcript_target_not_video.
When captions exist, data.transcript carries WebVTT. Check meta.creditsCharged on every response; it is the billing source of truth. Base cost is documented on the operation; AI fallback adds credits when you opt in.
When teams use it
Research agents, brand monitoring, RAG over creator clips, and sponsorship keyword scans need the words on the video, not a download-and-transcribe DIY stack. Homegrown downloaders fork into "parse VTT if present" and "run STT if not," with two failure modes and two bills.
Try a single URL in the free tool, then call the API with x-api-key for production jobs.
FAQ
How do I call the TikTok transcript API?
Call GET /v1/tiktok/videos/transcript with your Social Fetch API key (x-api-key) and a public video url query param. Read data.lookupStatus and data.transcript. Full parameters are in the API docs.
Does it always return captions?
No. Many clips have no caption track. Without useAiFallback, not_found means no text was available. With useAiFallback=true, Social Fetch can run speech-to-text for uncaptioned video at the documented credit cost.
What URL formats work?
Public TikTok video URLs, including /@user/video/… links and vm.tiktok.com short links. Photo carousels return transcript_target_not_video.
Is this the same shape as YouTube transcripts?
Both use data.lookupStatus and the shared success envelope. TikTok returns WebVTT (plus optional AI fallback). YouTube returns plainText plus timed segments and has no AI fallback path. Parse each platform's transcript object for the fields it documents.