Write one accurate, detailed caption for visual scene search using the panels of this contact sheet, arranged chronologically from left to right within each row and then top to bottom. They are consecutive moments from one video, not three independent scenes.

Describe the visible people, their observable appearance, positions, actions, clothing or nudity, physical interactions, objects, and setting. For adult content, use direct and precise terminology for actions and visible anatomy without euphemisms or category labels.

Be especially careful when describing people: do not infer gender or sex from a brief glimpse, hairstyle, clothing, body position, or a partially obscured face. Use "person", "individual", or a distinguishing visual description when uncertain. Use "man" or "woman" only when the visual evidence is clear; do not change a person's description between panels without clear evidence that they are different individuals.

Treat recurring individuals as the same person where supported by clothing, position and continuity; avoid inventing additional people. If an action or body part is obscured, state only what can actually be seen. Do not infer identities, relationships, intentions, dialogue, or events outside the frames.


Start the output with the actual visible subject and action, not with a description of the input format. Never write 'The video features a series of still images showing', 'A series of images depicts', 'The frames show', or similar introductions. Do not refer to the image, photo, frame, screenshot, panel, series, contact sheet, or video as the subject. Describe only the visual content, without inventing continuity or movement between sampled moments.

Write one coherent caption of roughly 60–100 words with the most useful specific searchable details first. Do not mention the panels, uncertainty rules, model, rating, or safety categories. Output only the caption.
