videodb

Index the video with video.index_spoken_words(force=True), then call video.search(query) with a natural language query. Results return shots with timestamps that you can compile into a playable stream URL. Wrap the call in try/except since it raises InvalidRequestError when no results are found.

Index spoken words first with video.index_spoken_words(force=True), then call video.add_subtitle() to burn subtitles into the stream. You can customize appearance with SubtitleStyle parameters like font name, size, and color.

Yes, video.reframe() converts aspect ratios server-side with presets like vertical (9:16), square (1:1), and landscape (16:9). Limit to short segments with start/end parameters or use callback_url for async processing, since reframe is slow on long videos.

No, desktop capture currently supports macOS only. The CaptureClient records screen, microphone, and system audio on macOS, streaming events over WebSocket. For other platforms, use RTSP/RTMP live stream ingestion instead.

The search method raises InvalidRequestError with 'No results found' when there are no matches. Catch this exception and treat it as an empty result set. For scene search, also set score_threshold to 0.3 or higher to filter low-relevance noise.

The timeline editor does not support transitions, speed changes, crop/zoom, color grading, picture-in-picture video, volume mixing, or keyframe animation. For those operations, fall back to local tools like ffmpeg as noted in the editor limitations.