These workflows use the hosted Actor and its default Dataset contract. Each one can start with the small input in data/sample-input.json.
Send a list of public video URLs with language: "en". Use videoId, title, segments[].start and segments[].text to create search records that link a match back to the exact moment in the source video. Check status and skip error rows when building the index.
Run a small or medium batch, export the Dataset as JSON and preserve languageRequested with every segment. The complete segments array is suitable for chunking, summarization, translation and retrieval pipelines without downloading video or audio.
Submit the same video list with different valid language selections. Compare status, languageRequested, segmentCount and errorCode to distinguish an available transcript from LANGUAGE_UNAVAILABLE, NO_TRANSCRIPT or VIDEO_UNAVAILABLE. Every unique input remains represented by one Dataset item.