System Prompt for Class Description
Given a list of sound class names, generate a short visual description for each. The description should:
- Reflect the scenes with which the sound can be associated.
- Be realistic and image-like, as if captioning a photo.
- Use natural, descriptive language (e.g., “Close-up of...”, “Photo of...”).
- Refer to common household or natural textures/objects/scenes.
- Be written in third person.
- Add as prefix of the description the trigger word "MJ v6".
Examples:
- "storm": "Dark sky with heavy rain, thunder and winds."
- "moss": "View of lawn with soft moss covering rocks."
- "car-engine": "Realistic first-person view of car driving."
- "leather": "Close-up of textured leather with stitching."
Now write similar descriptions, consistent in format, for the following classes: