Videos

Submit videos, list them, and read their scenes.

📄️Submit a video for indexing

Hand the indexing pipeline a video URL. The response is fast (202 Accepted) — the actual indexing runs asynchronously through scene detection → captioning → embedding → moderation. Poll `/v1/jobs/{job_id}` to track progress.

📄️List videos in your account

List videos owned by your account (org), newest first. Paginate with the opaque `cursor`: pass the `next_cursor` from the previous response; a null `next_cursor` means the last page.

📄️Upload one video file for indexing

Multipart-upload a single video file straight into the indexing pipeline. The same async flow as `POST /v1/videos`, but for a local file instead of a URL — poll `GET /v1/jobs/{job_id}` for progress.

📄️Get indexing result for a video

Return the full indexing result for an indexed video: AI summary, genre / IAB categories, GARM brand-safety assessment, per-scene captions + transcripts, and whole-video aggregations (brands, products, people, locations, audio events, mood timeline).

📄️Page through a video's scenes

Paginated, fully-hydrated scenes for a video — caption, transcript, on-screen text, mood (valence / arousal), detected brands / products / people / locations, plus presigned keyframe and clip URLs.

📄️Stream scenes as they finish analysis (NDJSON)

Push each scene's analysis as soon as its caption is ready, without waiting for the whole video to finish indexing. NDJSON like `/v1/ads/match` — one JSON object per line: