Good Use Cases for AI Image Generators

What is generative AI good for? From my experience as a product manager, technologist, and founder, I envision that image generation will be an important part of our future.

Before starting Tess, my cofounder and I previously worked on Google Image Search, so we experienced there how consumers browse and search for usable images. We also started an an online video editor called Kapwing, which has embedded AI image generation and media search. Now, I run Tess, where we’ve partnered with ~20 artists to launch a marketplace of visual styles. I've talked to dozens of customers and friends about Tess and how it might be helpful for them. Here are my learnings on how it might be used.

Not Good For

Overall, Generative AI makes images rather than "art." I envision that most art - high-impact, complex projects where the visual aesthetic is the central purpose - will still involve a direct commission between artists and publications.

The humans are the ones who bring a creative vision to the table, and a lifetime of studying, creating, and imagining art gives an art professional stronger vision for metaphors than a machine or an untrained person. Generative AI platforms like Tess may enrich conversations between illustrators and artists who can use the tools to brainstorm, mock up concepts, and bring ideas to life, but it won't replace art and design teams doing complex, high-visibility work. For example:

• Murals
• Magazine covers
• Wall art, galleries, and exhibits
• Interactive exhibits
• Mainstream commercial packaging

Because they're more sensitive to cultural context, artists are also better at incorporating modern trends and breaking news into their work than AI models. For similar reasons, people will be the ones pushing the experimental boundary of art, innovating on the frontier. Hopefully, generative AI will free up artists' time and mindspace for more avant-garde focus.

In addition, Generative AI is not yet good for precise illustrations like infographics, graphs, or comics. If you need exactly 5 people in the picture, most gen AI models have a hard time obeying instructions. Generative AI is also poor at crafting text and typography, although new models like Stable Diffusion 3 are improving.

Good For

Generative AI is especially helpful for projects with large output. Whereas these commissions may have been prohibitively difficult and time-intensive in the past, an artist can take on these commissions leveraging AI to quickly produce images, curate and tweak the outputs rather than manually draw each. For example, Tess customers and artists have said they use Generative AI for:

• Game design details, like the packaging of in-game wine bottles.
• Repetitive landing page icons
• Large image sets, like table placards or wine bottle labels
• Cover images for digital articles and social media
• Newsletter imagery
• Children’s book illustrations and cookbook art
• Individually-designed playing cards (see below for Queen of Hearts and King of Spades examples), tarot cards, or calendar pages

Generative AI also helps for series of art with similar visual themes and style, especially when the author needs a quick turnaround. I imagine that an artist may sell a license to a company to use their visual style rather than needing to draw each image; this is our business model at Tess. Some examples:

• Images for a daily column or op-eds
• Thumbnails for YouTube videos
• Art for podcast episodes or radio shows
• Album art or Spotify images for musicians

Finally, with a human artist at the wheel, generative AI can be helpful for brainstorming concepts and getting ideas. Tess can help open conversation between an artist and a client to express ideas visually:

• Set design, concept art, storyboarding, and costume and character design on TV shows or multimedia
• Impossible collaborations, like with a dead artist
• Unreal concepts, like “a dolphin on Mars”
• Tattoo designs
• Merch design ideas (to pint on hats, sweatshirts, tees, etc)
• Fashion and textures
• Wallpapers and backgrounds

Conclusion

Midjourney - a startup allegedly making $200M in revenue - has proven out the demand for generative AI, but unfortunately has kept this cash for themselves instead of sharing it with artists whose style they emulate. At Tess, we’ve launched an alternative business model that avoids copyright infringement and supports a healthy ownership ecosystem. If you’re an artist who wants to own your style, join us – we have about 20 artists onboard and are activating more accomplished illustrators.