2.05 Creating descriptive scene texts for visualisation and archiving

The value of descriptive text

In cultural heritage work, we often deal with memories of places, events, or objects that no longer exist or were never captured in photographs. A narrator’s spoken description is incredibly precious but can also be abstract.

Transforming that spoken memory into a rich, detailed, and evocative written text serves two vital purposes. First, it acts as a creative brief for an artist – be it a human illustrator or an AI image generator – providing the specific details, mood, and sensory information needed to create a faithful visual representation. Second, the descriptive text itself becomes a valuable archival object. It preserves the “mental image” of a place or event in a durable and shareable format, preserving not just the facts of a story, but its feeling, atmosphere, and soul.

AI demonstration: Generating a scene description

AI text generators are exceptionally skilled at taking key elements from a story and weaving them into a coherent and evocative scene description, enriching the provided information with descriptive language to create a compelling narrative paragraph.

Prompt:

Act as a descriptive writer. Based on the following key elements from a HER[AI]TAGE interview about a traditional festival on the river Mura in the 1950s, write a single, vivid paragraph describing the scene. The tone should be warm and nostalgic. Key elements: summer evening, bonfires on the riverbank, villagers gathered, playing traditional music on tamburica instruments, smell of grilling fish, reflection of fire on the water, wooden boats decorated with wildflowers.

Example AI response:

The warm summer evening air along the Mura was alive with the crackle of bonfires lining the riverbank, their flickering light dancing on the faces of the gathered villagers. The sweet, melodic strains of tamburica music drifted through the twilight, a joyful soundtrack to the laughter and conversation. A delicious aroma of freshly grilled fish hung in the air, mingling with the scent of woodsmoke. On the dark, glassy surface of the river, the firelight shimmered like scattered gold, reflecting the simple wooden boats that bobbed gently, their sides adorned with garlands of wildflowers picked that afternoon. It was a scene of simple, profound community, woven from light, music, and shared tradition.

AI demonstration: Refining the description

One of the most powerful aspects of working with an AI is the ability to iterate and refine the output. If the initial description is not quite right, a follow-up prompt can adjust the focus, add details, or change the mood. This conversational process allows you to co-create the text with the AI until it perfectly captures the desired feeling.

For instance, a follow-up prompt could ask the AI to rewrite the scene from the perspective of a child experiencing the festival for the first time, which would dramatically shift the tone and focus of the description.

Follow-up prompt:

That is a beautiful description. Now, please rewrite it to place more emphasis on the sounds of the scene and the feeling of community.

Example refined AI response:

The air of the summer evening along the Mura was not quiet but filled with a symphony of community. The primary sound was the joyful, rhythmic strumming of tamburica music, a melody that everyone knew by heart and that seemed to pull the whole village together. Underneath the music, a chorus of happy voices – excited children’s shouts, the low murmur of elders sharing stories, and bursts of laughter from all around – created a tapestry of human connection. The sharp crackle and pop of the bonfires provided a steady percussion, a sound that promised warmth and light as the sun dipped below the horizon. It was a feeling of belonging made audible, a moment where every sound was a note in the song of their shared life on the river.

Archival practice

These AI-generated descriptions are valuable assets and should be stored in your digital archive right alongside the original interview transcript and audio file, linked by the same unique identifier from your metadata schema. This ensures all related materials are kept together as a cohesive archival unit. These texts can then be provided directly to a human illustrator as a creative brief, used as the basis for a detailed prompt for an AI image generator, or read aloud as part of an audio guide for an exhibition.