# Parse Video

Transcribe speech from video into document records and text for summarization, extraction, and saving.

Use it to process spoken content in a training video. This node transcribes speech in the video. It does not interpret individual frames or describe shots.

## Configure transcription

Select an uploaded video file or bind the `sources` output from Read MOI Volume.

| Setting | Description |
| --- | --- |
| Language | Choose `auto` if unknown so the transcription service detects it |
| Noise reduction | Configuration default: `auto`; controls preprocessing before transcription |
| Minimum silence interval | Default: 0.5 seconds; range: 0.1–2 seconds; helps identify segment boundaries |
| Maximum speech duration | Default: 30 seconds; range: 5–60 seconds; limits each speech segment |

Values outside these ranges cause the node to fail. These durations apply to segments, not to the entire file.

## Use the output

Pass `documents` to document-processing nodes, or use `text` for summarization. The `text` field is omitted when empty. Segment content and metadata are carried in the document array.

## Example: summarize a training video

Flow: Read MOI Volume → Parse Video → AI Inference → Save to MOI Volume.

Bind `text` to the AI Inference prompt, ask for the topics and action items, and save the generated text.

## Related pages

- [Variables and data flow](../../variables-and-data.md)
- [Run and debug](../../run-debug.md)