Thirteen years ago, to complete an assignment for a journalism course, I wrote a humorous short story about a man who ate luxurious cat food.

This morning, I sat in front of my computer and witnessed a generative artificial intelligence platform called Lore Machine turn my novel into a comic.

I input my story into a text box, and the screen displayed the following message: "We are identifying scenes, locations, characters, and atmosphere. This process may take 2 minutes."

Lore Machine analyzed the text, extracted descriptions of characters and locations, and then handed this information over to the image generation model. Subsequently, a storyboard with illustrations popped up on the screen.

When I saw the characters from the story that I had almost forgotten displayed in vivid comic effects, my heart even skipped a beat.After more than a year of development, Lore Machine has been opened to the public for the first time. For a monthly subscription fee of $10, you can upload a total of 100,000 characters of text (up to 30,000 characters at a time) and generate 80 images for short stories, scripts, podcast transcriptions, and more.

Advertisement

It also has more advanced user plans, including a corporate plan for $160 per month, which can receive 2.24 million characters and generate 1,792 images.

These illustrations come in a range of preset styles, including comics, watercolors, and 1980s TV show styles, among others.

Lore Machine founder Tobey Campion demonstrated the capabilities of the tool to the founder of the creative agency Modern Arts, Zac Ryder, who has been using an early version of the tool ever since.

At that time, Ryder sent Campion a script for a short film, and Campion used Lore Machine to turn it into a 16-page comic novel overnight.Lei said: "I remember Campion shared his screen, and we were all completely stunned. It was not so much the image generation capability that impressed us, but rather the level of storytelling it achieved. From the narrative flow to the emotions of the characters, it did a great job."

Modern Arts is now using Lore Machine to develop a fictional universe for a comic series, the text of which comes from the creator of the popular Netflix series "Love, Death and Robots."

Behind the shocking performance, Lore Machine is actually composed of familiar components. A large language model scans your text, identifying descriptions of characters and places as well as its overall mood. Subsequently, it uses a version of Stable Diffusion to generate images.

What sets it apart is its ease of use. From uploading my story to downloading the comic storyboard, I clicked the mouse about six times.

This makes it one of the new batch of very easy-to-use artificial intelligence tools, which hide the amazing power of generative models behind a one-click web interface.The CEO of content creation company New Computer, Ben Palmer, stated: "To keep up with the development of new artificial intelligence tools, a lot of work is required, and each tool has a different interface and workflow.

Using a large tool with a unified UI design is very attractive. I think this is the direction the industry is heading."

No prompts needed

Two years ago, Campion established the company behind Lore Machine, dedicated to developing a blockchain-based Wikipedia. But when he saw how people used generative models, he changed direction.Using the free text-to-image model Midtravel, Campion created a comic version of the classic work "The Rime of the Ancient Mariner" by the British poet Samuel Taylor Coleridge. He said that these results went viral online, but they were not easy to produce.

"I woke up at four in the morning every day to work on these images," he said, "My wife hated that project." The problem is that text-to-image models like Midtravel only generate images one after another.

This makes it difficult to maintain consistency for the same character across different frames, and it is also hard to lock in a specific style across multiple images.

Campion said, "I eventually turned to a more convoluted and abstract way of expression."

This experience made him realize that this technology needs to be easier to use. Campion did not specifically explain how Lore Machine maintains the consistency of its images and style across a series of comic frames.The Lore Machine is quite impressive, but it's not perfect. In one scene of my story, a character with short hair grew bangs; in another scene, a character appeared twice.

These comic images also gradually began to feel somewhat generic, with the plot depiction not being specific enough. However, compared to manually and repeatedly entering prompts, this is a huge improvement.

"The consistency is very good," said Reid. It gives Modern Arts the confidence to use the Lore Machine in its clients' projects. He said, "If we had to keep going back to solve consistency issues, we wouldn't be able to deliver the images on time."

Like all generative models, the technology behind the Lore Machine can generate toxic content based on user instructions. Campion said that they have prevented it from generating images depicting violence or hateful stereotypes.

But beyond that, he is unwilling to suppress the creative expression of artists. For example, creating illustrations for celebrity fan fiction is allowed.Initially, the companies most interested in the Lore Machine were marketing agencies. However, Campion hopes that this public release will encourage more users to give it a try.

He said that six months ago, the principal of a school in Manhattan, USA, called him. The school is specifically designed to serve children with learning disabilities.

The principal wanted to use this tool to turn his textbooks into comics, so that children could learn by looking at the pictures. Campion said, "I never even thought of this, because I was too fixated on things like Hollywood."