Practical test: How ChatGPT can support media management

By teamnext Editorial Team

In the digital age, efficient media management is an important capability for companies. The professional term is Digital Asset Management, or DAM. Media content is treated as an asset. Digital Asset Management covers the processes required to store, organize, find, and distribute digital media such as photos and videos.

What role can an AI based chatbot play in a DAM context. The following sections explore this question using practical examples. The focus is ChatGPT by OpenAI.

Note: OpenAI is also known for the image generator DALL E.

How AI can improve DAM workflows

In many industries, AI has supported DAM workflows for years. A common example is the automated detection of people and objects in images, based on image recognition. ChatGPT can accelerate additional processes. These use cases are mainly related to language processing.

ChatGPT support for metadata optimization

ChatGPT is a language model trained on very large amounts of text. This textual knowledge can be used to design a metadata schema. The goal is a schema that is unambiguous and covers the categories that matter for a specific media library.

Side note: Additional metadata is essential for organizing digital media. It improves search and day to day file management significantly.

Practical test 1: Creating a metadata schema with ChatGPT

The first test asked ChatGPT, version 4, to design a metadata schema for a commercial media archive focused on sports photography, with an emphasis on Olympic sports.

Prompt:

  • Develop a detailed metadata schema for a commercial digital media archive whose content focus is sports photography, especially Olympic sports.

ChatGPT proposed 18 categories:

  • Image ID: unique identifier for each photo

  • Title: title of the photo. If not entered manually, it can be generated from other metadata elements

  • Description: short description of the captured moment

  • Keywords: relevant terms that support search when not covered by other fields

  • Date: the date the photo was taken

  • Location: the geographic location of the event

  • Creator: the photographer who took the photo

  • Source: where the photo originated, for example an agency or publication

  • Rights: copyright information and potential usage restrictions

  • Format: file type and file size

  • Resolution: pixel dimensions

  • Color mode: whether the photo is in color or grayscale

  • Event: the sports event, for example Olympic Games, world championships

  • Sport: the specific sport shown

  • Athlete(s): names of the athletes shown

  • Nation: the country represented

  • Medal winners: whether medal winners are shown and which medal applies

  • Medal ceremony: whether the photo shows a medal ceremony

Assessment

Categories 1 to 12 largely match common industry practice for professional media databases. Categories 13 to 18 are tailored to the defined use case, a sports photo archive. The schema is not final, but it is a strong baseline.

Takeaway

  • ChatGPT can help create an initial metadata schema that is aligned with a specific media library.

Practical test 2: Creating a keyword taxonomy with ChatGPT

Controlled vocabulary is essential for consistent tagging and metadata quality. For heterogeneous collections, a structured taxonomy, often including a synonym list, is critical. The more diverse the collection, the more dimensional the vocabulary structure must be.

The second test asked ChatGPT to create a keyword taxonomy for the same archive focused on Olympic sports. The output, in condensed form:

  1. Sport

  • Athletics

    • Track events

    • Field events

    • Combined events

  • Gymnastics

    • Artistic gymnastics

    • Rhythmic gymnastics

    • Trampoline

  • Aquatics

    • Swimming

    • Open water swimming

    • Artistic swimming

    • Diving

    • Water polo

  • additional sports

  1. Sports event

  • Olympic Games

    • Summer Olympic Games

    • Winter Olympic Games

  • World championships

  • Continental championships

  • National championships

  • additional event types

  1. Location

  • Countries

    • USA

    • Japan

    • France

    • additional countries

  • Cities

    • Tokyo

    • Paris

    • Los Angeles

    • additional host cities

  1. Athletes

  • list of athlete names, to be maintained and updated

  1. Scene

  • action shot

  • moment of victory

  • medal ceremony

  • crowd reaction

  • opening ceremony

  • closing ceremony

  • additional moments

  1. Result

  • gold medal

  • silver medal

  • bronze medal

  • world record

  • national record

  • personal best

Assessment

The taxonomy is relatively shallow due to the narrow topic focus. The top level categories are reasonable and usable as a starting point for a structured controlled vocabulary.

Takeaway

  • ChatGPT can support the initial creation of controlled vocabulary, for example by generating a first taxonomy for a specific archive.

Future use cases for ChatGPT in DAM

With proper integration, ChatGPT could support the automated handling of requests for media content. This requires text based metadata and a technical connection to the DAM system. Image recognition itself is not the core capability of ChatGPT.

Potential functions include:

  • triggering targeted media searches

  • compiling media selections

  • drafting and distributing updates about media assets

  • improving access to knowledge bases, for example through technical answers or user guidance

Conclusion

ChatGPT can support the setup and structuring of a media database. The most relevant contributions are first drafts for metadata schemas and controlled vocabulary structures. Additional use cases depend on data availability and system integration. Documentation related tasks in DAM can be partially automated. This reduces effort and can increase the value of marketing assets such as photos and videos.