TikTok trending video search API: ranked metadata for research

Sume's trending video search, POST /v1/trending-videos/search, returns ranked metadata about public TikTok videos for a brand, product, creator, or keyword query: watch URLs, author handles, metrics, and relevance scores. It is a paid research utility, not a generation model, and it does not download or mirror videos.

Every detail below comes from the Trending videos docs and describes the production API.

How do I search trending TikTok videos?

Send a query; production requires it. Every other field is optional. The request below asks for up to 10 TikTok videos from this week for a keyword, each with a lightweight summary.

curl -X POST "https://api.sume.com/v1/trending-videos/search" \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "cold brew",
    "platform": "tiktok",
    "window": "this-week",
    "limit": 10,
    "summary_mode": "metadata"
  }'

Which fields can I send?

What does each result contain?

The response lists ranked videos with public watch URLs, optional cover thumbnails, author handles, metrics, relevance scores, and optional lightweight summaries. It does not return raw TikTok video CDN URLs. In production, ranking uses a generic ranker with no extra relevance or view floor. The exact schema is in the live OpenAPI document.

A typical video entry carries:

• url: the canonical public TikTok watch URL.
• cover_url: an optional display thumbnail, not a downloadable video.
• description, created_at, and region.
• author.handle and author.nickname.
• metrics, relevance, and scores.
• summary, when summary_mode is not none.

What does trending search not do?

Know these limits before you build on it:

• It does not generate media. It is a research utility.
• It does not download or mirror videos, and it does not return downloadable source files for face swap (Beta) or captions today.
• It returns links, not files: url is a public watch URL and cover_url is a display thumbnail.
• It does not return transcripts yet. summary_mode: "transcript" returns metadata plus an unsupported warning.
• It searches TikTok only.

How does it fit into a video workflow?

The docs describe a two-step pattern: use trending search for research, then generate with Avatar or other generators using your own public HTTPS media inputs. For example, check what is trending for a product keyword, then script a talking avatar video or burn captions onto footage you own.

What does a search cost?

Each accepted call reserves and captures a fixed per-call amount of Sume usage, and summary_mode: "metadata" is included in that same per-call price. The docs say to confirm live pricing in GET /v1/catalog; see also API pricing and How Sume pricing works.