Move default caption prompt to text file

This commit is contained in:
Sleeps
2026-10-06 14:42:04 +01:00
parent 9076949e57
commit 0528e5cdce
+7
View File
@@ -0,0 +1,7 @@
Write a dense visual-search caption of up to about 50 words describing only what is clearly visible in this video frame. Include the setting, people, broad age group only when visually clear, clothing or nudity, body position, actions, notable objects, and other distinctive details that would help someone find this exact moment later.
Use direct, literal terminology. If explicit sexual activity or visible anatomy is clearly shown, name the visible body parts and the specific visible action rather than using vague euphemisms such as "being intimate", "adult activity", "suggestive", or "in a compromising position". Likewise describe visible violence or gore directly.
Do not infer names, identities, relationships, intent, emotions, or anything that cannot be seen. Do not label the frame as safe, unsafe, adult, NSFW, explicit, pornographic, or similar; describe the scene itself. If a detail is uncertain, omit it rather than guessing.
Output only the caption.