DEUTLI Aims to Tame AI Art Chaos with New Structured Prompt System

FRIBOURG, Switzerland – March 30, 2026 – Swiss startup DEUTLI today announced the open beta launch of a new platform designed to solve one of the most persistent frustrations in the rapidly evolving world of generative AI: the unpredictable nature of text-to-image creation. The V1 system, now live, replaces the familiar but often maddeningly vague text box with a structured, guided process, aiming to bring order and predictability to digital art generation.

For millions of users, from professional designers to curious hobbyists, the experience of using powerful AI image generators like Midjourney or Stable Diffusion is a double-edged sword. The technology offers boundless creative potential, yet harnessing it can feel like a game of chance. DEUTLI's system is built on the premise that the problem isn't the AI, but the interface between human intent and machine interpretation.

The Frustrating Art of the AI Prompt

The core challenge DEUTLI addresses is a phenomenon often described by users as "prompt sensitivity." Minor, seemingly insignificant changes in a text query—a misplaced comma, a different synonym, or a slight reordering of words—can result in wildly different images, shifting composition, style, and quality unpredictably. This forces many creators into a cycle of what one industry observer called "brute-force trial and error."

"You can get a masterpiece on the first try and a complete disaster on the second, using almost the same words," noted an anonymous AI artist on a popular online forum. This inconsistency makes it difficult to iterate on an idea or produce a series of images with a consistent aesthetic, a critical requirement for professional applications in design, architecture, and marketing.

Research into human-AI interaction confirms that free-form text is often insufficient for conveying nuanced visual ideas. AI models don't interpret language with human context; they process keywords and patterns without the ability to ask for clarification. This ambiguity forces the user to master a new, complex skill: prompt engineering. While a market for pre-made prompts has emerged on platforms like PromptBase and PromptHero, and optimizers like PromptPerfect offer to refine text, the fundamental challenge of translating a clear visual concept into effective machine language remains a significant barrier for many.

From Chaos to Control: A Deterministic Path

DEUTLI's answer is to move away from chaotic text input and introduce what it calls a "deterministic approach." Instead of presenting users with a dauntingly empty input field, the platform guides them through a step-by-step process. It prompts the author to clearly define the critical elements of their desired scene: the subject, its action, the location, and the overall atmosphere.

This modular construction algorithm is designed to provide the neural network with a structured, unambiguous set of instructions, allowing it to reproduce the user's original visual intent with far greater accuracy. The company states that the interface's logic is rooted in the principles of 500 years of visual arts history and is further informed by the project founder's 40 years of practical experience in professional photography and design.

"Instead of typing, you're building," explained a user testing a pre-release version. "It forces you to think like a photographer or a director, considering each component of the shot." The system also includes built-in presets that help ensure a professional visual standard, aiming to elevate the quality of outputs beyond the often generic look that can result from simple prompts.

A Universal Translator for Generative AI

A distinctive feature of the V1 release is its ability to simultaneously generate queries in three different "dialects." This addresses a major workflow hurdle for creators who use multiple AI platforms. The optimal syntax for a prompt can vary significantly between generative engines; a prompt that works perfectly in DALL-E 3 might produce a subpar result in Midjourney without specific parameter adjustments.

DEUTLI's system acts as a universal translator, taking the user's single structured input and converting it into optimized prompts tailored for the specific nuances of various leading generative engines. This feature abstracts away the need for users to learn and remember the unique command structures and keyword sensitivities of each platform, allowing them to focus on the creative vision rather than the technical implementation. This interoperability positions the tool not as another AI image generator, but as a central hub for crafting the instructions that power them.

Empowering Users and Redefining Interaction

Beyond improving predictability, DEUTLI is also focused on empowering users to own their creative process. In a notable departure from cloud-based systems where user data is often locked into a proprietary ecosystem, all created parameters and settings can be saved to a local file. The platform uses an open and officially documented application/vnd.deut+json format, giving users the ability to build their own independent, private library of "golden prompts" from their first day of use.

This feature is particularly appealing for professionals and studios concerned with intellectual property and workflow integration. It allows for the creation of a secure, reusable asset library that is not dependent on DEUTLI's servers, ensuring long-term access and control over one's creative toolkit.

To encourage broad testing and feedback, the company has made the open beta freely accessible. Users can perform five prompt improvements without creating an account, with the daily limit increasing to 20 after a free registration. Active testers can also request manual extensions of their daily limits.

By moving from unstructured language to a guided, visually-informed construction method, DEUTLI is part of a broader trend in AI interface design that seeks to make complex systems more intuitive and powerful. The project's official slogan, "Don't type. Snap it in," encapsulates this philosophy of rejecting routine in favor of clarity and control, potentially signaling a new chapter in how humans collaborate with creative AI.