Echoneo Evolution: From Detailed JSON Schema to Text Prompts and the Wizard Tool
1 min read
Echoneo Evolution: From JSON to Text Prompts and the New Wizard Tool
When starting the Echoneo project, our goal was clear: to systematically cross-pollinate the concepts and styles of 28 distinct art movements through AI collaboration, creating a vast 28x28 matrix of artworks. To guide this systematic exploration, we initially developed a highly detailed JSON prompt schema (v1).
The JSON Schema (v1): Initial Vision and Reality
Our aims with the initial JSON schema were specific (as previously detailed in our post "Echoneo: Prompt Schema Documentation (v1)"):
The Echoneo prompt schema was designed with several key goals in mind:
- Standardization: To provide a consistent structure for defining artistic intent across the entire 28x28 matrix of artworks.
- AI Interpretability: To structure information in a way that AI models can more reliably interpret distinct conceptual, stylistic, and compositional requirements.
- Semantic Clarity: To clearly separate the conceptual origin (the "what") from the stylistic execution (the "how") for each artwork.
- Creative Control: To offer detailed parameters for guiding the AI while still allowing room for emergent results.
- Modularity: To allow
conceptDefinition
andstyleDefinition
blocks to potentially be mixed and matched programmatically.
Using this schema, we produced the initial 56 baseline images for the echoneo-0-[Y]
(Prehistoric concept across styles) and echoneo-[X]-0
(various concepts in Prehistoric style) series. We gained valuable experience during this process. However, we realized that regardless of the detail, the specific parameters in the JSON schema (especially nuanced style settings, composition details, or sometimes color palettes) were not always implemented precisely or consistently by the AI image generation models available at the time (like DALL-E accessed via the ChatGPT interface). More often than not, many of our detailed definitions were overlooked, or the AI added its own interpretation. This situation limited the practical effectiveness of the granular control the schema aimed to provide.
Discovery: The Power of Simpler Text Prompts
Based on these observations, we decided to try a different approach: simpler, yet structured, text prompts. We divided the prompts into four main sections:
- Art Style: Core visual characteristics and techniques of the desired style.
- Scene & Technical Details: Atmosphere, camera view, lighting, and technical constraints (size, ratio, negatives).
- Concept: What the artwork is about – the core idea, theme, subject, narrative.
- Emotion target: The desired feeling or impact of the artwork.
Experiments with this structure led to a surprising conclusion: we discovered that with these text prompts, which are significantly more human-readable and editable, we could generate visuals of similar quality (and sometimes more consistent) to those produced with the detailed JSON schema. The text-based approach suggested that AI models (especially those interacted with via language models) might be more successful at processing natural language instructions for this type of creative task.
New Direction: Embracing Text Prompts and the Wizard Tool
Following this discovery, we decided to adopt the 4-part text prompt structure for the ongoing production of Echoneo artworks. This approach offers a better balance between practical production efficiency and creative guidance with current AI tools.
To allow everyone to experience this new approach and explore the cross-pollination possibilities inherent in Echoneo, we developed the /echoneo/wizard
tool!
What is the Echoneo Wizard?
- Interface: Features two sliders (0-27) representing the 28 art movements for Concept and Style selection, along with four text areas corresponding to the main prompt sections mentioned above.
- Functionality:
- When you select an art movement from the sliders, the Concept or Style definitions predefined for that movement automatically populate the relevant text areas.
- You can directly edit these auto-filled texts if you wish.
- By selecting option 28, 'Custom', on the sliders, you can write your own Concept or Style definitions entirely from scratch.
- The "Try on GPT" button takes the generated or edited 4-part text prompt and directs you to ChatGPT, allowing you to generate the visual immediately.
- Purpose: With this tool, you can access the prompts for the 784 potential combinations in the Echoneo project, review them, modify them, or combine predefined styles/concepts with your own imagination to create entirely new echoes.
Example: Crafting a Custom Epic Wizard Prompt
The real power of the Wizard tool comes alive when you move beyond the presets and define your own vision. Imagine wanting to generate an artwork not directly from art history, but from fantasy – specifically, an epic wizard, rendered in the dramatic Baroque Art Style. You could select "Custom" (or index 28) for the Concept slider, leave the Style slider on "Baroque" (index 9), and fill the text areas like this:
**Concept Definition:**Depict an immensely powerful, ancient wizard atop a storm-wracked mountain summit, silhouetted against tumultuous, divinely lit clouds. Visualize him actively commanding the tempest – raw lightning crackling visibly around his raised, ornate staff, wind tearing dramatically at his heavy, richly textured robes. Focus on the theatrical gesture, the intense, perhaps tormented or ecstatic, expression of concentration, and the dynamic interplay of arcane energy and chaotic natural forces. Capture a climactic moment of peak magical exertion, suggesting cosmic influence and immense power straining against its limits.**Emotion Target:**Evoke a profound sense of awe-inspiring power and grandeur, touching the sublime. The atmosphere must be charged with high drama, conflict, and intense, almost overwhelming energy. Capture the terrifying majesty of controlling nature's fury, the potential ecstasy or immense burden of wielding such cosmic forces, and a feeling that is epic, operatic, and deeply resonant with intense spiritual or elemental struggle.
Using these definitions for the Concept:
and Emotion target:
fields, along with the standard pre-filled text for Art Style: Baroque
and relevant Scene & Technical Details:
, you can hit "Try on GPT". This would generate a prompt designed to produce an image echoing the specific vision described – like the artwork shown on the top.
This demonstrates how the Wizard tool facilitates not only the exploration of art history combinations but also the creation of entirely new artistic narratives using the defined stylistic frameworks.
Connecting Past and Present
While the shift from JSON schema to text prompts might seem like a deviation from our initial vision, it's actually an adaptation more suited to the project's spirit and the current capabilities of AI tools. The fundamental research and definition work we did analyzing art movements in depth for the conceptDefinition
and styleDefinition
remains valid and valuable – the Wizard tool now presents this structured information in a more accessible and interactive format.
This evolution demonstrates the dynamic nature of creative collaboration with AI; adapting to the tools and their current capabilities is necessary to achieve the best results.
You are invited to try the Echoneo Wizard and create your own Echoneo echoes!