I tried every AI design tool. Here are the results.

Summary

In this video, Parvin reviews 23 AI design tools, comparing their outputs using the same prompt. The prompt focused on creating a clean, minimal, and elegant e-commerce website using natural earthy tones, high-quality craft packaging, and specific fonts. Tools like Magic Patterns, Lovable, Mocha, and Firebase Studio stood out for their creativity and adherence to guidelines, while others like Google Stitch and Durable fell short. The final ratings were based on design quality, color scheme adherence, font usage, and responsiveness. The video offers valuable insights for digital designers and AI tool enthusiasts.

Takeaways

• 😀 Parvin tested 23 different AI design tools over four weeks, including unreleased ones, to compare their outputs and find the best.
• 😀 The tools were tested based on a fair, consistent prompt that focused on design features like high-quality craft packaging, minimalism, and using earth tones.
• 😀 Some design tools performed well by interpreting the prompt correctly, including Magic Patterns, Lovable, Mocha, and Firebase Studio.
• 😀 Magic Patterns stood out for its creativity in using a circle placeholder image and a step-by-step CTA but lacked perfect font and image placements.
• 😀 Lovable received praise for its innovative scrolling card feature but faced issues with odd layouts, number-icon placements, and secondary buttons.
• 😀 Mocha impressed with its beige and amber palette, excellent pricing card design, and overall cohesive feel, though it still had room for improvement in detail.
• 😀 Firebase Studio was noted for its consistent spacing, good type scale, and overall polished output, making it a favorite among testers.
• 😀 Base 44 was appreciated for its clean design and consistent use of a circle motif, even though it didn't perfectly follow font and color guidelines.
• 😀 Tools like Subframe, Onlook, and Durable failed to meet expectations due to poor color handling, unresponsiveness, and lack of section variety.
• 😀 Google Stitch was especially disappointing for not following the prompt, with issues like misalignment, incorrect fonts, and using AI avatars that didn't fit the design.

Q & A

What was the primary goal of the experiment described in the script?

The primary goal was to test and compare various AI design tools to determine which one performs best in terms of generating high-quality design outputs, particularly in the context of a DTOC (direct-to-consumer) e-commerce brand with specific design requirements.

How did the creator ensure the fairness of the experiment?

The creator used the same design prompt for all AI tools, ensuring consistency across the experiment. This allowed for an apples-to-apples comparison between the tools, and excluded those that didn’t have a oneshot text-to-prompt interface.

What specific design guidelines were included in the prompt?

The prompt asked the AI tools to create a design for a DTOC e-commerce brand with craft cardboard packaging, using natural earthy tones, minimal clean design with generous white space, and an elegant yet exciting aesthetic. Specific fonts (Instrument Sif and Space Mono) were also provided, and placeholders were requested instead of real images.

What was the rationale behind excluding some AI tools like Cursor and Wind Surfer?

The tools were excluded because they didn’t have a oneshot text-to-prompt interface, which was a key criterion for ensuring a consistent comparison. The creator wanted to compare tools that operated under the same framework.

Which AI tools performed exceptionally well in the experiment?

Lovable, Mocha, and V0 tied for first place, as they produced high-quality design outputs that aligned with the prompt’s guidelines. Firebase Studio also performed well and was highly rated by one of the designer friends, Henry.

Which AI tools were rated poorly and why?

Vizily, Durable, and Google Stitch were rated poorly. These tools failed to adhere to the guidelines, with issues like not following the color scheme, misusing fonts, and not providing responsive versions of the designs. Google Stitch, in particular, did not meet the prompt’s requirements and failed to generate usable code.

What was a standout feature of Magic Patterns that impressed the creator?

The standout feature of Magic Patterns was its unique and non-standard approach to the design, such as using a thick border around the placeholder image and offering creative step-by-step guidance in the CTA section, which contributed to a potentially high-converting landing page.

What problems were observed with Firebase Studio despite its positive feedback?

Firebase Studio had some issues, such as footer padding problems and minor design imperfections. However, overall, it was praised for its consistent spacing, type scale, and good icon usage, making it one of the top-rated tools.

Why did the creator have high expectations for Subframe, and what went wrong?

Subframe was expected to perform well due to its similarity to Figma, making it closer to a design tool rather than just a oneshot prompt generator. However, it failed to execute the color scheme properly, misused fonts, and lacked responsiveness, which significantly impacted its rating.

What did the creator mean by 'system prompts' and how did they affect the AI tools?

System prompts refer to the predefined instructions that many AI tools rely on to generate design outputs. The creator suspected that many tools were limited by these system prompts, which resulted in generic, templated designs that didn’t fully deviate from the standard structure.