TikTok Transcript API

Use cases

• Analyze what creators say at scale
• Track brand and product mentions in videos
• Feed short-form video into LLM pipelines
• Caption and translate TikTok content

Example request

Send a GET request with your key in the x-api-key header.

curl -G 'https://staging.socialtap.dev/v1/tiktok/transcript' \
  --data-urlencode 'url=https://www.tiktok.com/@niinoclock/video/7673108027645414678' \
  -H "x-api-key: $SOCIALTAP_API_KEY"

const params = new URLSearchParams({
  url: "https://www.tiktok.com/@niinoclock/video/7673108027645414678",
});
const res = await fetch(`https://staging.socialtap.dev/v1/tiktok/transcript?${params}`, {
  headers: { "x-api-key": process.env.SOCIALTAP_API_KEY },
});
const { data, meta } = await res.json();

import os
import requests

res = requests.get(
    "https://staging.socialtap.dev/v1/tiktok/transcript",
    params={"url": "https://www.tiktok.com/@niinoclock/video/7673108027645414678"},
    headers={"x-api-key": os.environ["SOCIALTAP_API_KEY"]},
)
body = res.json()
data, meta = body["data"], body["meta"]

Parameters

Every endpoint also accepts max_age.

Example response

A 200 response is {data, meta}: data is a transcript object, meta describes the request (meta fields).

{
  "data": {
    "available_languages": [
      "en"
    ],
    "duration_ms": 495000,
    "id": "7673108027645414678",
    "is_auto_generated": false,
    "language": "en",
    "platform": "tiktok",
    "segments": [
      {
        "end_ms": 4820,
        "start_ms": 100,
        "text": "Let me show you how to make a Japanese cream stew!"
      },
      {
        "end_ms": 8380,
        "start_ms": 4821,
        "text": "I think there is a way to definitely make it from scratch,"
      },
      {
        "end_ms": 13120,
        "start_ms": 8381,
        "text": "but most Japanese people buy the stock for the cream stew."
      }
    ],
    "source": "captions",
    "text": "Let me show you how to make a Japanese cream stew! I think there is a way to definitely make it from scratch, but most Japanese people buy the stock for the cream stew. If you go to a Japanese supermarket, you can find this next to or near where the curry stock cubes are. This is…",
    "url": "https://www.tiktok.com/@niinoclock/video/7673108027645414678"
  },
  "meta": {
    "request_id": "req_3f9c2a7d1e8b4c6a9f0e2d4b6a8c0e1f",
    "platform": "tiktok",
    "source": "captions",
    "is_cached": false,
    "credits_charged": 1,
    "latency_ms": 668,
    "fetched_at": "2026-10-25T12:00:00Z"
  }
}

Response fields

data follows the shared Transcript schema, identical on every platform.

Pricing

You pay only for successful requests. The price depends on which source produced the result, reported as meta.source; sources are tried cheapest first, in the order Captions → Speech-to-text.

Example: a result from Speech-to-text asr with 3 started minutes of media costs 2 + 2 × 3 = 8 credits.

Your balance must be at least 1 credit to start a request; with less, it is refused with insufficient_credits (HTTP 402) and costs nothing. Only that minimum is set aside when the request starts. One request costs at most 122 credits, but that amount is never locked up in advance: right before costlier work, SocialTap sets aside more of your balance, the exact price when the number of started minutes of media is known beforehand, otherwise as much as your balance allows (and the work is limited to what that pays for). It then charges the actual price and releases the rest. If the result would cost more than your balance covers, the request is refused with insufficient_credits as well, and costs nothing.

meta.credits_charged and the x-credits-charged header show what you paid. How credits work · Packs

Errors

Errors cost nothing. Codes marked “Yes, later” under Retry can succeed on a later attempt; the others return the same answer every time.

not_found, private_or_unavailable and no_transcript are item-level: the request was fine, the item can't be returned.

Measured performance

p50 latency

668 ms

p95 latency

2,705 ms

Success rate

99.1%

Requests measured

112

Staging measurements, not production. Measured on our staging environment, over the 7 days before 2026-10-25. How it is measured.

FAQ

Related

More endpoints returning transcripts