Douyin Video Search API

Douyin Video Search

Features

• One-field search contractSubmit only query. The endpoint rejects undeclared inputs such as results_limit and does not require a provider token or internal identifier.
• Bounded source requestEvery task asks the provider for page 1 at offset 0, without exposing provider pagination controls to the caller.
• Normalized reel-video recordsRead available text, author, media, duration, timestamps, and visible play, like, comment, and share counts through a consistent result schema.
• Asynchronous stored-result readsSubmit once, poll the task, and use SocQ's result cursor only when the completed task stored more records than one response page returns.

Parameters

How to Use

Provide a non-empty query, wait for the asynchronous task, and read the normalized records stored for that task.

1. Choose a specific text query that describes the public videos you want to discover.
2. POST query to /v1/douyin/video-search and save the returned task_id.
3. Poll /v1/tasks/{task_id} until the task succeeds or reports a failure.
4. Read data.results.items. If has_more is true, pass next_cursor to the task-result request to continue through stored SocQ records.

curl -X POST "https://api.socq.ai/v1/douyin/video-search" \
  -H "Authorization: Bearer $SOCQ_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query":"street food"}'

# Poll GET /v1/tasks/{task_id}
# Read data.results.items; next_cursor pages stored SocQ results only

Best Use Cases

• Topic-led video discovery: Create a bounded snapshot of public videos returned for one product, theme, event, or research phrase.
• Creator-context review: Connect returned captions and media with the available public author fields for qualitative research and dataset enrichment.
• Format and media analysis: Review available video media, thumbnails, dimensions, and duration alongside the source record and query context.
• Visible engagement snapshots: Compare available play, like, comment, and share counts while preserving collected_at as the observation time.

Pro Tips

• Require meaningful non-whitespace query text in your own interface before submitting the task.
• Do not send results_limit, page, offset, search_id, or a provider cursor; they are not part of the public SocQ request schema.
• Treat the returned set as one provider response, not an exhaustive search index or a caller-controlled ranking window.
• Use GET /v1/tasks/{task_id} cursor and limit only for stored-result reads; limit defaults to 50 and accepts 1 through 500.
• Keep collected_at and handle optional author, media, publication, duration, and individual metric fields without substituting zero for missing values.

Related APIs