OpenArt: Into the future of personalized artificial intelligence
This year, artificial intelligence and image generation flourished. I have been following a new entrant in this field — OpenArt. OpenArt is a new platform for “AI art enthusiasts” that is building infrastructure to create, discover, and monetize AI-generated art.
Last month, OpenArt launched PhotoBooth. Using research published by Google and Boston University, PhotoBooth creates personalized AI models so users can generate images of themselves or their pets. Prior to it, text-to-image models were unable to create images of unique subjects because they are trained to synthesize 5+ billion photos from across the web. With this research, AI can now mimic specific people, pets, and objects when given a handful of photographs. PhotoBooth enables anyone to pay a nominal fee to access this technology.
Creating a text-to-image model of my dog
I used OpenArt’s PhotoBooth feature to create a text-to-image model of my dog, Skye. It prompted me to submit a collection of photographs showing Skye in different poses, backgrounds, and lighting (I submitted 24 images).
Next, OpenArt guided me in naming the model (#Skye#) and the class (dog) so that I could reference it in future text-to-image prompts. The model took 40 minutes to train. Once complete, I received an initial 64 images that were generated with preset prompts. (Someone wrote the PhotoBooth prompt package I purchased and receives revenue share each time it is bought.) The package includes eight prompts that rendered Skye as a cartoon firefighter, a Disney character, an underwater diver, and so on.
OpenArt also provides 400 credits with each transaction, allowing users to generate images using their own prompts. Knowing how to communicate with text-to-image models helps creators to leverage AI. OpenArt’s Prompt Book (November 2022) introduces text-to-image technology and guides readers in basic prompt engineering. The Prompt Book answers questions including: Can you tell the model to prioritize certain attributes using capitalization or word order? How do text-to-image models interpret subjective words? Can I create images with a distinct artistic style?
I set off to experiment with PhotoBooth. I opted for the modifier “Baroque” in order to create a regal portrait of my dog. Baroque refers to a style of European art, music, and architecture that is highly embellished and realistic. Compared to other images generated with Stable Diffusion, this image is ornamented (see the gold crown, collar, and background) and realistic (high resolution and shadows). Baroque is just one example; vocabulary can be borrowed across art mediums, artists, and movements.
Next, I prompted my text-to-image model to create images in the style of Monet. Monet was an avid painter of nature and a founding figure of modernism. His paintings have visible brushstrokes, muted colours, and honest depictions of the outdoors. The AI-generated images reflect these key characteristics.
Photography prompts are equally effective. I referenced photographic techniques including close-up, telephoto lens, and portrait style to generate the images above. Prior to PhotoBooth, I would have needed specialized knowledge and equipment to create these pet portraits. Instead, being literate in prompt engineering enables me to create art across a variety of mediums and styles.
The future of personalized AI
OpenArt is an early leader in the AI-art space. The platform is a new way to socialize and connect, while making breakthrough AI research accessible to large audiences.
It is especially impactful for students and early-career creatives, who can experiment with low costs.
I see a future where young architects and designers generate proof-of-concept imagery using AI to secure new opportunities to create in our physical world. Authors will illustrate their stories with AI-generated art to tell stories that resonate across language and culture. For those outside the creative industries, platforms like OpenArt will ensure anyone with an internet connection can shape the future of artificial intelligence in our world.