Compare these two camera states and decide if there's a meaningful change.

PREVIOUS: "{prev_summary}"
CURRENT: "{response}"

If the situation is essentially the SAME (same person, same place, same activity):
→ Respond exactly: NO_CHANGE

If something CHANGED (person appeared/left, moved, different action):
→ Respond with short summary: "Person: [what changed]"

Examples:
- "Person at desk" vs "Seated person at computer" → NO_CHANGE
- "Person at desk" vs "Person walking to door" → Person: walking to door
- "No person" vs "Person detected" → Person: appeared

Respond with ONLY "NO_CHANGE" or a short summary:
