Extract video frames for AI and ML training data

JAD divides the video duration by the total number of cells in the grid and seeks to each interval point. For a 4×4 grid (16 cells) on a 4-minute video, frames are sampled every 15 seconds. The first cell is at position 0 (or 1 frame offset to avoid a black opening frame) and the last cell is near the end of the video. The sampling is even across the full duration.

Yes — enable the timestamp overlay option to add the source timecode (HH:MM:SS) in the bottom-left corner of each frame. This is particularly useful for review and logging purposes — when viewing the contact sheet, each frame can be referenced back to the exact source timecode for editing decisions.

The frame grid resolution depends on the source video resolution and the number of cells. For a 1920×1080 source with a 4×4 grid, each cell is 480×270 pixels — the output grid is 1920×1080 pixels total. For a 4K source, cells are proportionally larger and the grid can reach very high resolutions. A maximum output dimension cap (default: 8000×8000px) prevents excessively large outputs.