Douyin Transcripts Scraper — Subtitles & Translation

Extract spoken content from public Douyin (抖音) videos into original text, timestamped cues and subtitle files. Supply video URLs, mobile share links, share text or numeric video IDs. Use it for China-market content research, subtitle editing, multilingual review and AI analysis of videos you are authorized to process.

Private candidate. End-to-end source and translated delivery has been observed. Broader release acceptance, comparative accuracy and target pricing remain under validation. This Actor is not ready for Store publication.

What you receive

One dataset row per delivered video transcript, with full JSON and TXT exports. SRT and WebVTT exports are enabled by default with no separate export event. Optional translation retains original speech and adds translated text and aligned subtitle files.

Field	Meaning
transcript, segments	Original spoken text and cues with start/end seconds
translatedTranscript, translatedSegments	Optional translation with the same cue boundaries
transcriptSource, timestampSource	Native captions or speech recognition; model timings are estimates
transcriptComplete, translationStatus	Explicit completeness and translation outcome
jsonUrl, txtUrl, srtUrl, vttUrl	Full transcript and subtitle export links
words	Optional speech-model word timings; native captions have cue timing only
awemeId, videoUrl, scrapedAt	Identity, canonical source URL and UTC collection time
Caption, engagement, creator details	Available public context, separate from spoken words

For public metadata without speech, pair this Actor with Douyin Video Scraper. For audience discussion, pair this Actor with Douyin Comments Scraper. Use this Actor when you already have public video links and need spoken text. Do not use this Actor for keyword discovery or creator-profile collection.

How to get Douyin transcripts

Add public video links, choose original speech or a translation, set duration and event budgets, then run. Read both the dataset and RUN_SUMMARY; use each row's export links for full files.

{
  "videoUrls": ["https://www.douyin.com/video/7534679152504376595"],
  "maxVideos": 10,
  "targetLanguage": "original",
  "exportSubtitles": true,
  "maxTotalMinutes": 30,
  "maxChargeUsd": 5
}

Desktop /video/ links, iesdouyin.com/share/video/ links, v.douyin.com links, feed links with modal_id, pasted share text and bare numeric IDs are accepted. Up to 100 inputs are accepted; maxVideos defaults to 10. Canonical duplicates are removed before delivery and billing.

Input options

Option	Default	Scope
transcriptMode	auto	Native captions then speech; speech regenerates audio text; captions uses existing tracks only
targetLanguage	original	Original speech or one of 51 configured translation targets
includeWordTimestamps	false	Word timing on the speech-recognition route
exportSubtitles	true	SRT and WebVTT exports; TXT/JSON always retained
responseFormat	detailed	concise gives 1500-character and ten-cue previews with full export links
maxVideos	10	1–100 videos per batch
maxDurationMinutes	10	Universal 1–30 minute limit per video
maxTotalMinutes	30	1–300 rounded audio minutes per run
maxChargeUsd	5	Transcript and translation event budget; startup additional; platform usage paid by the creator
maxRunSeconds	900	60–3600 second execution deadline

The Actor never presents post captions, hashtags, music titles or reusable music tracks as spoken transcripts. Unknown counts remain null. The lightweight public page can omit comments, shares, saves, music and follower counts; metadataUnavailableFields identifies those gaps and metadataSourceUrl records the page used.

Translation and subtitle workflow

{
  "videoUrls": ["7534679152504376595"],
  "targetLanguage": "en",
  "transcriptMode": "speech",
  "includeWordTimestamps": true,
  "exportSubtitles": true
}

The form offers 51 translation targets, including English, Arabic, Spanish, Hindi, Japanese and Korean. These are implemented settings, not individually verified language-quality claims. Original speech is always retained. Translation must preserve every cue index. Incomplete output is rejected; failed translation preserves the source transcript, reports partial delivery and incurs no translation event.

A researcher reviewing Mandarin product demonstrations can collect source speech, add English captions for colleagues and keep both languages together. An editor can import SRT or VTT into a compatible editing tool and check timings against the video. This workflow does not verify claims made by the speaker.

Output example

This shortened example illustrates the contract; it is not a cloud-run sample. Full rows include metadata and export links.

{
  "awemeId": "7534679152504376595",
  "videoUrl": "https://www.douyin.com/video/7534679152504376595",
  "durationSeconds": 67.756,
  "status": "success",
  "transcript": "你好，欢迎观看。",
  "segments": [{"start": 0, "end": 2.5, "text": "你好，欢迎观看。"}],
  "transcriptSource": "speech-recognition",
  "timestampSource": "model-estimated",
  "transcriptComplete": true,
  "targetLanguage": "en",
  "translatedTranscript": "Hello, welcome.",
  "translatedSegments": [{"start": 0, "end": 2.5, "text": "Hello, welcome."}],
  "translationStatus": "complete",
  "billedMinutes": 2,
  "responseTruncated": false
}

Pricing and cost controls

Private candidate pricing is Pay per event with creator-funded platform usage, with equal features and event rates on every plan. Check the live Pricing tab after private setup and before running.

Event	Proposed price
Actor start	$0.00005 per allocated GB, minimum one event
Delivered source transcript	$0.019 per started audio minute
Successful optional translation	Additional $0.009 per started audio minute

Exports have no separate event charge. A 67.756-second video rounds to two minutes: $0.038 for source text, or $0.056 with translation. At default 1 GB memory, startup adds $0.00005. Platform usage is paid by the creator; verify the effective Pricing tab before running. The $5 per 1,000 short-video target is not an active launch price. A batch with 5 results from five one-minute videos costs $0.095 in source-transcript events before startup.

maxChargeUsd excludes startup and platform usage. Apify's own maximum-total-charge setting also limits events. Videos exceeding the duration limit are skipped rather than silently shortened. Processing is sequential to bound memory and costs. A $2 owner-provider reservation limit per run includes failed requests; a provider budget stop can leave work unprocessed.

No transcript row or value charge is created for known unavailable videos, no audio, no recognized speech, a slideshow or missing captions in captions mode. The startup event may still apply; platform usage is paid by the creator. Files are saved before row delivery and billing. Interrupted delivery stops automatic recharging and requires inspection of saved files and event charges.

API example

After private deployment and approval of cloud tests, use your own Apify token. The response contains run and storage identifiers; poll completion and retrieve the dataset and summary.

curl --request POST \
  --header "Authorization: Bearer $APIFY_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{"videoUrls":["7534679152504376595"],"maxVideos":1,"maxChargeUsd":0.10}' \
  'https://api.apify.com/v2/acts/khadinakbar~douyin-transcripts-scraper/runs?maxTotalChargeUsd=0.11'

Use defaultDatasetId with Apify's dataset API. Export rows as JSON, CSV or Excel through Apify storage tools. Subtitle files are separate key-value-store records. Access and retention follow storage settings; links are not promised to last forever. No cross-user cache is used.

AI-agent prompt

Use khadinakbar/douyin-transcripts-scraper for spoken text from these public Douyin URLs. Retain source speech, request English only when needed and cap transcript events at $1. Read run outcome and dataset before summarizing. Report incomplete text, failed translation, provenance and export links. Use separate Douyin discovery tools to find videos.

Direct API use is intended. Hosted MCP discovery, exposed schemas, stateless execution and payment eligibility remain unverified for this private candidate.

Best results and troubleshooting

Begin with one short, clearly spoken video you can currently open publicly. Use a direct URL if a stale share link fails. Choose speech to regenerate absent or unsuitable captions. Dialects, singing and noisy music can reduce recognition quality. Verify important quotes and editing timings against the source.

VALID_EMPTY records known skip reasons. INVALID_INPUT includes a correction hint. PARTIAL preserves results and reports failed or remaining work. UPSTREAM_FAILED and CONFIG_ERROR with no useful output honestly fail. Retry target-access failures later; configuration failures require owner attention. Reduce the batch or adjust universal duration budgets for limit-related skips. Reconcile interrupted billing instead of blindly replaying it.

Builder's note and verification methodology

Post captions and music tracks are easy to confuse with speech. The adapters separate them, validate video identity, label transcript provenance, preserve null counts and verify translation cue alignment. More fields alone would not prove better transcript quality.

Research checked on 2026-10-14 against live Apify Actor pricing metadata found competitors with lower prices and others offering MP4 archiving. This candidate targets batch input, preserved source text, provenance, inclusive subtitle exports and plan-neutral limits. Cloud delivery is under active validation; same-job price and reliability targets are not yet met. No universal superiority claim is made.

Public-data boundaries

Process public videos you are authorized to collect, transcribe and translate. Review platform terms, copyright, privacy and applicable law for your use. Restricted/private videos and credential-based login are outside scope. Temporary media is deleted after processing; transcript files remain in your run storage under its retention settings. Owner provider secrets are supplied through encrypted environment configuration and excluded from results and logs. This independent tool is not affiliated with Douyin, ByteDance or competing developers.