Custom Prompts for AI-generated Descriptions

Sculpt Your Image Metadata with DBGallery’s Custom Prompts

Introduction

Managing and tagging thousands of digital assets can be an overwhelming task. Whether you are a professional photographer, a marketing team, or an e-commerce business, ensuring your images have accurate and searchable metadata is crucial. DBGallery’s new custom AI prompt feature simplifies this process by allowing users to generate precise descriptions and relevant tags with just a few words.

How Custom AI Prompts Work

This innovative feature lets users enter their own prompts, guiding AI to create metadata tailored to their needs. Instead of relying on generic auto-tagging, DBGallery’s AI can produce industry-specific descriptions, thematic tags, creative captions, or even structured data based on the ability to enter a custom AI prompt. This level of customization enhances the efficiency, accuracy and value in digital asset management.

Custom prompt for ai-generated descriptions.

Use-cases for custom prompts

Three use-cases will be covered here:

  1. General Descriptions: Great prompts asking for a general description.
  2. Structured data prompts: Request data in a required format. This might be to feed to other systems, such as an ERP for example.
  3. Generate tags only: No wordy description, just the metadata tags.

 

Use Case 1: General Descriptions

A. E-commerce Product Listing

Prompt: “Generate a product description and SEO tags for this clothing item.”

Output: “A stylish blue denim jacket with a classic cut, perfect for casual or semi-formal wear. Tags: denim, blue, fashion, outerwear, trendy.”

B. Creative and Marketing Team Social Media Post

The AI can generate compelling product descriptions for branding and marketing campaigns.

Prompt: “Create an engaging social media caption and hashtags for this image.”

Output: “Make a statement with bold colors! 🎨 #CreativeDesign #Photography #BoldColors”

C. Professional Photography Archive

Prompt: “Describe this image in detail, including lighting, composition, and mood.”

Output: “A golden-hour landscape shot with warm lighting, dramatic shadows, and a serene lake reflecting the mountains.”

D. Real Estate Property Description

Prompt: “Provide a property description suitable for real estate listings, emphasizing key features.”

Output: “A modern 3-bedroom home with an open-concept kitchen, hardwood floors, and a spacious backyard, ideal for families.”

E. Personalized Search Enhancements

Generate context-aware asset descriptions based on user preferences.

Prompt: "Describe this image for an advertising agency looking for urban lifestyle visuals."

Output: "A dynamic and energetic city scene featuring a diverse group of young professionals walking through a bustling downtown district. The warm golden-hour lighting enhances the modern glass skyscrapers in the background, while street vendors and cyclists add to the vibrancy of the urban environment. The image conveys a sense of ambition, connection, and contemporary city life—ideal for advertising campaigns targeting young professionals, tech startups, or lifestyle brands."

Use Case 2: Structured data prompts

Sometimes the AI-generated data is used for more than just searching within DBGallery.  Get more than just descriptions of an image, get valuable data that can be used in other systems...or possibly for special purposes within DBGallery, like structured searches.

Exporting data: Often useful for this type of generated image data, DBGallery has a CSV data export capability.  This allows structured description data to be exported for consumption by internal or 3rd party systems.   

NOTE: Explore what can be outputted, tweaking the prompt as many times as needed.  The custom prompt, and the resulting data, is extremely flexible.  The output format and data possibilities is the same as can be returned from the latest and best AI's.  As of March 2025, DBGallery uses ChatGPT o1-mini's API for generating descriptions. 

While the number of use cases for this flexible capability is nearly endless, three scenarios are described here to provide food for thought on the possibilities.

Scenario 1: Structured data to feed to a 3rd-party system, such as an ERP system

Prompt: This is a fashion photo.  Describe the type of material and major colors of the top and bottom clothes worn.  Do so in 4 sentences, with a new line after each, in the format "Top Color: <colors>" "Top  Material: <material>" "Bottom color: <colors>" "Bottom Material: <Material>.

Result (on the right): 

custom prompt fashion example.

 

Scenario 2: Auto-prepared Website Data

Prompt: "Provide a structured SEO-friendly metadata description, including a title, meta description, and alt text for this image.  Provide it in html format. "

Molecular sample photo.

Result: 

<!DOCTYPE html>
<html lang="en">
<head>
    <meta charset="UTF-8">
    <meta name="viewport" content="width=device-width, initial-scale=1.0">
    <title>Antibody Drug Conjugate Visualization</title>
    <meta name="description" content="Detailed illustration of an antibody drug conjugate (ADC), showing the antibody, linker, and payload components in action against target cells. Perfect for educational and research purposes in biochemistry and pharmacology." />
</head>
<body>
    <img src="path-to-your-image.jpg" 
         alt="Illustration of an Antibody Drug Conjugate (ADC) showing the Antibody, Linker, and Payload components interacting with a target cell" />
</body>
</html> 

 

Scenario 3: Identifying Damage in Inspection Images:

A property inspection company uses a Digital Asset Management (DAM) system to store images of buildings, vehicles, or equipment. The goal is to detect and tag damage based on predefined categories (e.g., cracks, rust, dents, leaks) while starting the output with a clear "Yes" or "No" regarding damage presence.

Prompt:

"Analyze this image and determine if damage is present based on the predefined list: [Crack, Rust, Dents, Water Damage, Mold, Scratches]. If damage exists, list detected types; otherwise, respond 'No damage detected.'"

Sample AI-Generated Output (With Damage):

Yes, damage detected.
Damage Types: Crack, Water Damage, Mold

Sample AI-Generated Output (No Damage):

No damage detected.

Use Case 3: Generate tags only

Similar to prompts for structured data output, here your company's unique objects are recognized and tagged within a set of tags provided in the prompt.  It is like having a custom-trained AI for object recognition.

Scenario:

A manufacturing company stores images of its products in a DAM system. Each product has a predefined list of possible tags (e.g., part numbers, serial codes, material types). Instead of generating a full description, they need only the tags present in the image for accurate cataloging.

Prompt:

"Identify and extract only the tags from this image based on the predefined list: [Steel, Aluminum, Plastic, Part#1234, Part#5678, Serial#A001, Serial#B002]. Ignore any other text or objects."

Sample AI-Generated Output:

Tags detected: Steel, Part#1234, Serial#A001

 

Tip: Use an AI to suggest a custom prompt for your scenario

It can be challenging to come up with the more effective prompt for some images.  For example, to describe your company logos you could simple prompt with "Describe these logos".  When asking ChatGPT for a suggested prompt to it come up with one that is far better.

We asked ChatGPT: "What would be a good ai prompt for describing company logos?"

Response:

You are an expert brand designer and marketer. I will provide the name of a company and its logo (or a brief visual reference). 
    Your task is to describe the logo in a structured, detailed way. 

Please follow these rules in your response:
1. Identify the company name and logo text (if any).
2. Provide a concise description of the logo’s overall design (shapes, main elements, colors, layout).
3. Mention any significant symbols or icons within the logo and what they might represent.
4. Note the primary color scheme (hex or simple color terms).
5. Include any relevant brand attributes or style descriptors (e.g., minimalistic, bold, playful, professional).
6. If the logo includes a tagline, mention it and explain how it relates to the design.

Format your response as follows, with a blank line between each:
Company Name: 
Visible Text: 
Overall Layout/Shape: 
Color Palette: 
Main Symbol or Icon (if any): 
Brand Impression or Style: 
Tagline (if applicable): 
Additional Observations:

Use clear, professional language, and stay objective while noting possible symbolic meaning or brand associations.

In another case, we asked ChatGPT: "We have a number of second hand item photos that needs to identify one main item, not just described.  What prompt would you suggest?"

Response:

You are an expert product identification model. Your goal is to identify the single main item in the photo, not just describe it. 
Follow these steps: 1. **Identify the primary item**: Determine exactly what the item is (e.g., “Dell laptop,” “Adidas running shoes,” “ceramic vase,” etc.). 2. **Give key details**: - Category or item type (e.g., “Laptop,” “Shoes,” “Furniture,” etc.). - Brand or manufacturer (if any logo, label, or distinctive brand feature is visible). - Color and material (if identifiable). - Any unique or notable feature (e.g., model number, style, condition). 3. **Ignore background or secondary objects**: Focus only on the main item that someone would sell or buy. 4. **Output** the information in a structured format. For example: Category: [item category] Brand: [brand name or "Unknown"] Color/Material: [color(s), material(s)] Main Identifier: [a concise name or model if known] Key Features: [any visible unique elements or condition notes] If you cannot determine a specific detail (like brand), write “Unknown” or “Not visible.” Important: Do not include disclaimers or extra commentary. Simply provide the identification details based on the image.

Output within DBGallery:

The Value of Custom Prompts

  • Saves time by automating metadata generation.
  • Enhances searchability by ensuring consistent and relevant tags.
  • Improves discoverability for businesses using AI-generated metadata for SEO and marketing.
  • Increases flexibility by allowing metadata to be tailored to specific industries and use cases.
  • Structured data can be used for a variety of business uses.

 

Try it today!

To test drive any of the above, click the Try it Free button at the top of this page for a 15-day trail.  No credit card required.  Just upload your images then use the custom prompt feature, which has easy to understand instructions in our Knowledge Base: Custom Prompts for AI-generated DescriptionsBe sure to experiment with all the possibilities for your business.  Alternatively, contact us for a free and eye-opening demonstration.

Conclusion

With DBGallery’s AI-powered custom prompts, managing digital assets has never been easier. Whether you need structured metadata to support business processes or creative captions for social media, this feature allows you to define exactly what you need.

DBGallery Logo Spinner