Describe what you want to see. INKBOT Lite helps you explain it to an AI image generator.
Have a picture in mind but do not know how to write the prompt? Start with ordinary words—even a rough description. INKBOT Lite helps organize your subject, actions, objects, and important details into a structured visual instruction.
Copy the finished instruction into an external image-capable AI to create the picture. Review the result, identify what needs changing, and refine your idea.
INKBOT prepares the instruction. An external AI generates the image. Access requirements and image-generation limits depend on the service you choose.
You do not need to begin with polished prompt-writing language. Put an ordinary idea into the workspace, let INKBOT help structure the visual handoff, and give that instruction to an external image-capable AI.
A lightweight workspace designed around one simple question: can we get the idea onto the page clearly enough for a person to recognize what the AI understood?
Begin with the picture you mean—not a collection of technical keywords. Identify what matters, prepare the handoff, and use the result as a starting point for correction and refinement.
View INKBOT Lite in the ShopStart by explaining what the picture should show. You do not need technical language. Identify the people or objects, what is happening, where it happens, and which details must stay accurate. INKBOT Lite helps organize that description into an instruction for an external image-capable AI.
Describe the subject, action, setting, and visual style. Explain important relationships, such as who is holding an object or where each person stands. For example: “A simple black-line drawing of an adjuster and homeowner reviewing damage photographs across a table, on a white background.”
Yes. Begin with the words you would use to explain the picture to another person. INKBOT Lite is designed to help organize rough descriptions into a structured visual handoff. You then give that handoff to an external image-capable AI.
Compare the result with your original description. Identify the specific mismatch: the wrong subject, a missing object, an incorrect position, or the wrong drawing style. State what should change and what should remain unchanged in your next instruction.
No. INKBOT Lite prepares the visual instruction. An external image-capable AI generates the picture. Access requirements, image limits, and available features depend on the external service you use.
See how to prepare your first AI image prompt with INKBOT Lite
INKBOT is organized around a recognition loop rather than a one-shot prompt. The first image is not treated as the final answer. It lets you see what the system understood, identify what is right or wrong, and continue.
This diagram describes the broader architecture under exploration. The important boundary is between what the human means and what the AI infers. INKBOT is designed to make that boundary visible rather than hiding it inside a single prompt.
The goal is not to replace a capable generative model. The goal is to improve the information and decision layer surrounding it. These components describe the broader INKBOT approach.
Start with ordinary language. Rough words, fragments, incomplete thoughts, and voice-dictated ideas can become the authoritative starting record.
Organize the idea into subjects, actions, attributes, relationships, spatial arrangements, scale, sequence, and purpose without unnecessarily replacing the original meaning.
The broader architecture explores preserving relevant project context and retrieving useful references while distinguishing known information, remembered information, and new inference.
Check whether the visual instruction still represents the important parts of the original idea before sending it to the external generative system.
The first picture provides a visible interpretation of the idea. It creates something you can inspect rather than evaluating a prompt in isolation.
Keep what is right and identify what needs to change. The next instruction can become more precise because you have something concrete to react to.
A deliberately complicated physical-world concept was used to explore whether an ordinary-language idea could be translated into a visual engineering representation.
The concept involved a technician walking a beach with a survey wheel and downward-facing smartphone camera, automatically capturing overlapping images, recording GPS coordinates, and stitching the sequence into a continuous beach walk.
A remote viewer would select a shell from the imagery, and the associated location would guide the technician back for physical retrieval.
The visual storyboard made the proposed physical arrangement, image sequence, overlap, coordinates, remote selection, and retrieval workflow visible enough to inspect and correct. It illustrated the concept rather than establishing that the complete physical system had been built or validated.
Explore the Architecture Behind the ExperimentThe deeper INKBOT architecture explores how memory, references, semantic parsing, validation, provenance, synchronization, external-model instruction, visual correction, and revision state can work together without making the human give up control of the original idea.
The deeper edition is intentionally model-agnostic: INKBOT constructs the instruction architecture and hands the visual-generation step to an external AI system.
View INKBOT Editions in the ShopINKBOT Systems is an independent concept laboratory exploring visual thinking, human–AI interaction, instruction architecture, and practical concept development.
INKBOT is an independent prototype and concept laboratory. It is not an OpenAI product, partnership, certification, endorsement, or representation of OpenAI.
The project is intended to make the human–AI instruction problem visible through working prototypes, experiments, and architectural exploration.
INKBOT prepares instructions for external AI systems. It does not guarantee that a service will generate an image, follow every instruction, or reproduce an idea exactly.