Video Intelligence

Interlocute supports common video formats including MP4, WebM, and MOV. Videos can be submitted via URL fetch, blob storage upload, or direct upload depending on your use case.

Speech extracts transcripts and captions. Visual detects shots, scenes, OCR text, labels, and objects. Insights extracts entities, topics, sentiment, emotion, audio events, and safety signals. You can combine any or all of them in a single request.

When you select multiple profiles, the platform takes the union of their signal sets and selects the appropriate provider preset automatically. You are billed for the combined analysis, not per-profile — so combining Speech + Insights costs less than running them separately.

Yes, via the Insights profile. The index includes time-bounded sentiment segments (Positive, Negative, Neutral) and emotion segments (Joy, Sadness, Anger, Fear, Surprise, Disgust).

You request a textual summary specifying length, style, and optional custom instructions. The summary is generated asynchronously — you submit the request, poll for completion, and retrieve the result. Keyframe images can be included for visual context.

Yes. Video indexing operations are metered by duration and complexity. Every submission and result retrieval is logged in your usage ledger with full attribution to the node and API key.

Yes. The ReIndex operation lets you re-process an existing video with a different profile combination — for example, adding Insights to a video that was originally indexed with Speech only — without re-uploading the source file.