RabbleRouse News Exclusive: AI for Artists no. 5

This week’s column focuses on one of the most essential skills for artists working with Large Language models (LLMs) or any multimedia generation models: prompt engineering. Whether you’re generating images, drafting project descriptions, designing workshops, or brainstorming new ideas, the quality of your prompt can directly shape the usefulness and creativity of the output. For artists navigating an increasingly AI-driven creative landscape, understanding how to write effective prompts is a valuable skill.

Consider the difference between the two prompts:

Weak prompt:

“Write something about my art for my website.”

LLM response:

“You are an artist who makes art. Your work is creative and expressive. You are passionate about what you do.”

This response is vague and generic. Now compare that to:

Strong prompt:

“Act as a professional arts writer. Write a 150-word website bio for a performance artist whose work explores memory, diaspora, and ritual through movement, poetry, and installation. Keep the tone clear and accessible.”

LLM response:

A multidisciplinary performance artist whose work centers on memory, ritual, and migration, they create immersive experiences that utilize movement, spoken word, and installation-based practices. Influenced by diasporic cultural frameworks, their work invites reflection on identity, place, and belonging. Their performances have been presented across the U.S., offering audiences intimate encounters with layered histories and contemporary experience.

This comparison illustrates how prompt engineering can transform a model from a vague assistant into a creative collaborator. But first,

What is Prompt Engineering?

Prompt engineering is the art and science of crafting clear written instructions, known as prompts, to guide AI models, especially large language models (LLMs) and other generative AI models, for generating specific and useful results. It involves asking clear, context-rich questions that help the model understand your needs.

AI researcher and OpenAI co-founder Andrej Karpathy describes prompting as “programming in natural (human) language.” He recommends treating the model like a “very smart but extremely literal intern.” In this view, a well-constructed prompt must define the task, offer relevant context, and specify the desired format.

For example, instead of saying:

“Write something about my dance project.”

Try:
“You are a grant writer. Write a 150-word summary of a contemporary dance piece titled Echoes in Stillness that explores intergenerational memory through dance and spoken word. The summary should be suitable for an arts grant application.”

This revised prompt defines a role, sets constraints (such as word count and purpose), and provides context. In this case, ‘constraints’ refer to specific requirements or limitations that guide the format and content of the AI’s response.

A Simple Prompt Template

Here is a reliable template that can be used by artists to create their own prompts:

You are a [role]. Perform [task]. Include [context]. Follow [constraints].

Where: 

• Role: Who is the AI acting as?
• Task: What do you want it to do?
• Context: What is this about?
• Constraints: What style, format, tone, or length should it follow?

For example: 

“You are a communications assistant. Write a 100-word website bio for Mira Sinha, a multidisciplinary artist who works with movement, sculpture, and poetry. Highlight her interest in ecology and ritual. Use third person and a professional tone.”

However, note that prompting is rarely a one-time event. It is an iterative process. If the initial output isn’t quite right, tweak the prompt. Be more specific. Add examples. Break a complex task into smaller parts. With each revision, you clarify your intent and improve the results.

Few-Shot Prompting

Another powerful technique to improve LLM output is “few-shot prompting,” where you provide examples of the desired responses. This is especially useful when tone, structure, or format is essential.

For example, instead of: “Write a 200-word description of an art project about labor and urban space,” you might say: “Here are two sample project descriptions [Insert the two example project descriptions]. Write a third in the same tone and format for a new performance piece about memory and migration.”

This helps the model align its output with your desired style.

Unified Use Case: Mira Sinha

To illustrate how prompt engineering supports real creative tasks, let’s follow a single fictional artist: Mira Sinha, a multidisciplinary artist who works with dance, sculpture, and poetry to explore ecology, ritual, and memory. Using a consistent subject across examples demonstrates how refining prompts can yield varied yet coherent outputs.

1. Grant Application Summary
Example Prompt: You are a professional grant writer. Write a 200-word summary of Mira Sinha’s upcoming performance piece Seasons of Return. The piece blends Odissi dance and sculptural installation to explore cycles of ecological loss and rebirth. The tone should be accessible and suitable for an arts grant application.

2. Website Bio
Example Prompt: You are a communications assistant. Write a 100-word bio for Mira Sinha, an artist working with dance, sculpture, and poetry. Emphasize her engagement with ritual and ecological themes. Use a formal tone and the third person.

3. Creative Brainstorming
Example Prompt: You are a creative collaborator. Suggest five titles and concepts for future projects by Mira Sinha that expand on her interest in seasonal change and ancestral memory. For each, include a one-line concept and emotional tone.

4. Workshop Design
Example Prompt: You are a teaching artist. Design a four-week beginner-level workshop for emerging artists based on Mira Sinha’s project Seasons of Return. The theme is “Movement and Ritual.” Include weekly objectives, sample activities, and required materials. Format should be clear and ready to copy into a teaching document.

Prompting Against Homogeneity

As more artists adopt AI tools, a growing concern is that the convenience of default prompts will lead to an “AI aesthetic,” characterized by a flattened sameness in tone, structure, or visual style. But prompt engineering is the antidote to this. By providing detailed and personalized prompts, artists can steer models away from generic outputs and achieve distinct, meaningful results.

Specificity is key. You can reference lesser-known artistic movements, fuse disparate ideas, or set precise emotional tones. This type of intentional prompting resists homogenization and may facilitate a more effective synthesis of the desired outcome (whether it be AI-generated art or AI-generated script).

Tools and the Expanding Role of Prompts

Beyond LLMs, prompt engineering is also central to emerging creative tools. Saga is an AI platform supporting filmmakers from scriptwriting to previsualization. Runway’s Gen-2 and OpenAI’s Sora enable users to generate short video clips simply by describing what they want to see in words. For example, a user might input a prompt like, “a rainy night in Tokyo, filmed from a moving taxi,” and the tool will produce a short, animated video reflecting that vision. These tools require precise prompting for better results. Runway’s results improve significantly when prompts include specific styles, emotional cues, or camera angles.

From Tool to Practice

Prompt engineering is no longer a niche technical skill. It is a disciplined practice, a way of collaborating with AI models through language. By treating the prompt as a creative script, artists can direct these tools with greater control and clarity. The more refined your prompting becomes, the more your AI outputs will reflect your unique voice. In this way, prompt engineering becomes a way of preserving authorship in the age of generative AI.

Next time you open your favorite Gen-AI tool, don’t just ask. Prompt with intent.

Disclaimer: Portions of this column were developed with the assistance of generative AI tools, including ChatGPT and Gemini, which were used to help structure the content and identify illustrative examples. All interpretations and editorial decisions remain my own.