Claude Design : le guide complet pour les non designers (2026)

This video provides a comprehensive guide to using Cloud Design, focusing on its capabilities, potential pitfalls, and the crucial role of design systems in creating unique and differentiated digital products. The presenter, Shubam Charma, emphasizes that while AI tools like Cloud Design are powerful, their effectiveness hinges on how they are utilized. The core problem addressed is the emergence of "AI slop" – generic, uninspired designs that result from basic prompts and a lack of strategic approach. Cloud Design, a relatively new tool, has generated significant buzz, even impacting competitors like Figma. However, it's presented as a tool, not a magic bullet. The tutorial begins with navigating the Cloud Design interface, which is exclusively web-based. Users are guided to `cloud.ai` and then to the design section. The interface is divided into two main areas: a left panel for project management and creation, and a right panel for the design canvas. Projects can be initiated in two ways: as prototypes for applications, mockups, or landing pages, or as "slides" for presentations, banners, or static images. The presenter opts to demonstrate the "prototype" functionality first, focusing on creating an app for habit tracking. A key decision point is choosing between "wireframe" and "high fidelity." The recommendation is to start with wireframes, analogous to creating a house blueprint before decorating. This iterative approach helps avoid overlooking crucial steps and ensures the final result aligns with expectations. In the wireframing stage, users provide a prompt describing their desired application. Cloud Design then asks clarifying questions to refine the concept. For the habit tracker app, questions concerned the platform (mobile/desktop), types of habits, gamification style, and notification systems. The presenter deliberately provides a less precise prompt to illustrate how Cloud Design generates follow-up questions. Cloud Design then generates multiple design directions based on the prompt and answers. These are presented as mockups, allowing users to visualize potential outcomes. Users can then select preferred directions and provide feedback. The process is demonstrated with a habit tracker app, showcasing initial wireframe generation for both mobile and desktop. The presenter highlights how this visual feedback loop helps validate design choices early on. Another wireframe example is presented: an app for tracking video posting frequency, demonstrating variations in layout and control elements. Once a wireframe direction is chosen, the next step is to move to "high fidelity" design. The presenter advises against directly generating high-fidelity designs from the initial prompt. Instead, they recommend using the selected wireframes to generate a prompt for a high-fidelity version. This is crucial for avoiding "AI slop." By first creating wireframes, the AI is guided towards a more refined and less generic output for the high-fidelity stage. Screenshots of the chosen wireframes are uploaded, and the generated prompt is used to create a single, polished high-fidelity version of the app. The high-fidelity design process showcases Cloud Design's advanced editing capabilities. Features like "tweak" allow for quick iteration on elements like color palettes, offering multiple variations for selection. The "edit" function provides granular control over typography, color, and spacing, drawing comparisons to Figma's advanced features. The tool's ability to generate alternative components, such as replacing an Apple-like system with a custom design, is also demonstrated. The presenter also touches upon the collaboration feature, allowing users to add notes for team members. The distinction between Cloud Design and other AI tools like Lovable, Bolt, or V0 is clarified. While those tools aim for functional applications directly from a prompt, Cloud Design is positioned as a prototyping tool that generates designs which then need to be coded, often using tools like Cloud Code. This distinction is important because, according to the presenter, tools that aim for a full functional app in one go often produce results that fall into the "AI slop" category due to a lack of differentiation. The concept of "AI slop" is further elaborated: it's not about the technical quality or animations, but the lack of uniqueness. As AI lowers the barrier to entry for creating digital products, differentiation becomes paramount. The presenter admits to falling into this trap by using a single prompt for promotional materials, resulting in designs that are likely to become ubiquitous. This is attributed to the AI's tendency to adopt a default "house style" if not sufficiently guided. This leads to the central theme of the video: the design system. A design system is described as a "design grammar" that ensures consistency and uniqueness across all digital touchpoints, from websites to applications. It defines elements like typography, color palettes, spacing, and components. Renowned companies achieve brand recognition and consistency through their carefully crafted design systems. Examples like Starbucks and The Verge are cited to illustrate this. Cloud Design incorporates a dedicated feature for creating and managing design systems. Users can either import an existing design system or have Cloud Design generate one based on a detailed prompt. The presenter demonstrates the creation of a design system for a fictional company selling pool robots. The process involves defining brand style, colors, and other visual elements. While the initial generation might require some refinement, it provides a solid foundation. The presenter acknowledges that building a robust design system is a time-consuming process, even with AI assistance. The generated design system includes elements like logos, spacing guidelines, and button styles. These systems are downloadable and can be integrated into various projects, even with other tools. The presenter shares a personal example of using a pre-existing design system to generate a landing page for a workshop, showcasing how the AI leverages the system for consistent branding. Finally, the video covers publishing the created designs. Two methods are presented: a simple method for static web pages and a more complex approach for applications. For simple web pages, the design can be exported as a ZIP file containing HTML and other assets. This ZIP can then be uploaded to a web hosting provider like Hostinger. The key is to ensure the main file is named `index.html`. For more complex applications, the workflow involves "handoff" to Cloud Code, pushing to GitHub, and utilizing deployment tools like DocPloy. This allows for more sophisticated deployments. The presenter concludes by reiterating the importance of investing time in design systems, even when using AI tools. While Cloud Design can rapidly generate these systems, they require careful configuration and refinement. A well-defined design system is presented as the key to avoiding "AI slop," achieving differentiation, and maintaining brand consistency. The video also briefly touches upon best practices for prompting Cloud Design, emphasizing the need to provide context, objectives, layout, content, and audience information. This foundational prompting technique is applicable across various AI tools, including Cloud Code and Cloud AI, as Cloud Design is essentially an interface built on top of the Opus 4.7 LLM.