YouTube Transcripts API

YouTube Transcripts

Features

• Multi-video transcript retrievalRetrieve available transcript text from multiple public video URLs in one request.
• Preferred language selectionRequest a preferred language with a valid BCP-47 code such as en or zh-CN.
• Language availability metadataReturn the actual selected language and the list of available transcript languages.
• Plaintext transcript formatProvide a consistent plaintext format with transcript source information when known.
• Traceable video contextConnect every transcript with its video ID, title, channel information, publication date, duration, and collected_at.

Parameters

How to Use

Request captions only after choosing the videos that matter. Then verify the returned language and source before the text enters search, RAG, summarization, translation, or compliance workflows.

1. Choose public video URLs, ideally after discovery or metadata review has removed irrelevant items.
2. POST urls and an optional BCP-47 language to /v1/youtube/transcripts and store the task_id.
3. Poll /v1/tasks/{task_id}; a successful batch can still contain fewer results than submitted URLs.
4. For every item, inspect language, available_languages, and transcript_source before chunking the plaintext; paginate while has_more is true.

curl -X POST "https://api.socq.ai/v1/youtube/transcripts" \
  -H "Authorization: Bearer $SOCQ_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"urls":["https://www.youtube.com/watch?v=arj7oStGLkU"],"language":"en"}'

# Poll GET /v1/tasks/{task_id}
# Read GET /v1/tasks/{task_id}?limit=50

Best Use Cases

• Video summarization: Use available transcript text to create summaries, research notes, or content briefs for long videos.
• Searchable video archives: Index transcript text with video IDs, titles, channel information, and language metadata.
• Topic and information extraction: Identify topics, entities, claims, and keywords across a selected collection of public videos.
• Translation and localization: Use the selected and available language fields to route transcript text into translation workflows.

Pro Tips

• Always route by the returned language, not the requested preference. Fallback can legitimately select another available track.
• Machine-generated captions can contain names, punctuation, and domain-term errors. Keep transcript_source and review high-stakes extracts against the video.
• A successful task with fewer items usually means some videos had no usable public captions; it is not automatically a batch failure.
• Plaintext has no segment timestamps or speaker labels. Do not manufacture precise quote timing from this response.
• Chunk text downstream, but keep video ID, language, source, and collected_at on every chunk so citations remain traceable.
• Use Videos to enrich title, media, and public metrics when the transcript will appear in a searchable catalog.
• If exact language coverage is mandatory, compare available_languages first and treat fallback results as exceptions requiring review.

Related APIs