Inkitt, a self-publishing platform that uses AI to develop bestsellers, records $37 million led by Khosla.

The startup’s eponymous app lets users self-publish their stories, uses AI and data science to select and fine-tune the ones they find most compelling, and then Distribute and sell with the second app Galatea.

The company says its business already has 33 million users and dozens of best-sellers. The new Series C funding will be used to expand the types of content the company produces. AI writes stories based on users’ original ideas, creating personalized versions of fiction for specific readers. Transition to games and audiobooks. There is also a growing amount of video content based on fiction being published on the platform. These videos are currently produced by humans, but eventually they will be generated using AI.

CEO and founder Ali Albazaz said the long-term vision is to expand the content library and build a multimedia empire around it. Rather than a competitor like Wattpad, it’s more of what he calls the “Disney of the 21st century.”

The $37 million Series C is led by Vinod Khosla of Khosla Ventures. Previous backers NEA and Kleiner Perkins and other private investors also participated.

This investment brings Inkitt’s total raised to date to $117 million (other rounds include $3.9 million Seed, $16 million Series A, and $59 million Series B). Masu). The company has already received an acquisition offer from at least one publisher, and while consumer-centric startups rarely raise money, if Inkitt chooses to raise money, investors have already lined up the company for another round. We understand that we are approaching This current Series C values ​​Inkut at approximately $400 million post-money.

Inkitt’s ambition is, in some ways, to change the zigzag of the publishing industry.

Reading has been on the decline for the past two decades, due to the rise of mobile phones and apps designed to take up minutes (and sometimes more) of consumers’ leisure time. In 2022, the most recent study by the Bureau of Labor Statistics, the average American adult spent just 15 minutes a day reading, down from more than 20 minutes a few years ago. The average number of books read (completed) also decreased to about 5 books.

Inkkit is ironically trying to counter that trend by leaning into it. The company is focusing on more innovative ways to present books, building chapters that are shorter and easier to read on mobile devices, and using many effects (for example, as audio) throughout the text to enhance the reading experience. make it more dynamic.

It is also possible by tailoring the book to the reader’s preferences. Run A/B tests on every aspect of your work, from titles and story arcs to first lines and cliffhangers. This gives Inkkit a wealth of data about how fiction in general, not just specific books, is performing and what performs best. We found that the data applied more directly to subsequent books.

All of that seems to be working. Some services that rose to prominence during the height of the COVID-19 pandemic (such as “instant” delivery, online shopping, virtual events, and all sorts of streaming) have declined in the years since, or at least adopted a more moderate growth pattern. I’m back. That’s not the case for Inkitt, Galatea, and the startup’s newest company, Galatea TV, all of which are expected to see an increase in engagement time.

“People have so much going on right now,” Albazas said of the post-pandemic years. “They’re looking for an escape, and we think that’s why we’re thriving.”

He said that Inkitt is ranked as the 11th best-selling publisher in the world in total, surpassing well-known publishers such as Penguin Random House, and that the company’s algorithms improve the success rate of publishing best-selling books compared to traditional publishers. It claimed to be “20 times” more expensive than the publisher. .

Revenue doubled last year (although the company did not disclose actual numbers). Also, new projects like Galatea TV are an outgrowth of the reading business, but are also highly profitable in their own right.TV series based on the book Galatea beautiful mistake That alone brought in $500,000 in revenue, he said. The company is “almost profitable,” he added.

Founded in Berlin and now based in San Francisco, Inkitt focuses on fiction, although it’s not entirely accurate to say it aims for literary fiction. In fact, after a well-known author who dabbled with the platform a few years ago walked away in horror and disgust after Inkkit suggested some edits for an upcoming novel, Albazaz said Inkkit hopes to avoid headaches by working with big-name authors. Big egos focus on the long tail of undiscovered talent.

It will be interesting to see how much that long tail bends to the will of the algorithm. Albazaz said AI-generated stories, based on catchy treatments created by talented humans, are part of what he hopes to do more of in the future, as well as story personalization using AI algorithms. I said that there is.

Personalization remains a work in progress, and Inkitt is experimenting with different degrees of changes it might implement in the original work, or even potentially handing control over to the readers themselves, he added. Ta.

To realize this vision, the company is experimenting with a number of LLMs, including APIs from OpenAI, Anthropic, and Mistral AI (Albazzaz says the startup’s current favorites are suited for short passages) to build narratives. He said he is building his own customizations around them. He said.

All books are automatically published in 10 languages. DeepL is the primary AI used for this purpose. The company uses his Eleven Labs text-to-speech audio generator for audiobooks and is currently using Leonardo for cover art (he said it recently switched from Midjourney). The main idea seems to be to stick to mixing and triangulating to get the best results.

“These LLMs write very poor or average content,” he said. “Importantly, the LLM alone cannot create best-selling content, so we leverage the data we have collected over the past few years.”

Interestingly, there are also intellectual property reasons behind this. Inkitt doesn’t want to rely on a single his LLM to create a complete novel. Because even though the platform claims it doesn’t intend to use its own data to train its models, Inkitt doesn’t want that. Take a chance, he said. (To that end, to the LLM he sometimes very deliberately asks for monotonous prompts so as not to tell too much about Inkitt’s best work.)