Typeface is a platform trained on ChatGPT and Stable Diffusion models that can generate personalized blogs, Instagram posts and websites for companies.
Abhay Parasnis, the former chief technology officer of Adobe, wants to use AI powered by OpenAI, Stable Diffusion and computer vision models to help companies churn out branded content. His startup Typeface, which he launched in June 2022, has now raised $65 million in Series A financing to continue building out its generative AI platform for marketing and communication content such as blog posts, Instagram posts, websites and job postings on LinkedIn.
One of the challenges businesses face, Parasnis says, is that the skills needed for marketing can take years to acquire, which can make finding the right people challenging.
“The world where I come from, you need to master Photoshop for a decade before you can really use it to produce amazing content,” Parasnis tells Forbes.
Enter Typeface: The startup lets companies upload their existing content such as web pages, blogs, Instagram posts, brand logos and other visual assets. This personalized data set, combined with public data, is used to train Typeface’s model, which is built on OpenAI’s GPT-3.5 and a customized version of Stable Diffusion 2.0. The platform is designed to learn based on the company-specific data and create text and image content that is personalized to each enterprise’s brand voice and audience.
The funding round, which Typeface announced Monday, includes investors Lightspeed Venture Partners, GV (formerly Google Ventures), M12 (formerly Microsoft Ventures) and Menlo Ventures.
The presence of two AI heavyweights is telling: “Microsoft and Google are the two names that are … in the big AI fight, if you will, and Typeface is the one company where they both feel like it’s actually a unique enough perspective on enterprise that they both wanted to be part of the round,” Parasnis says.
VCs were ready to invest before Typeface was looking to raise funds, he says, largely due to the caliber of the talent on the startup’s team that includes engineers who worked on the AI-powered coding assistant GitHub CoPilot, Microsoft Azure and Adobe. Typeface, which has filed patents for its products, did not disclose the number of employees on its team.
Parasnis says Typeface says one of Typeface’s early customers is Sequoia Benefits Group, which uses the platform to create thousands of job descriptions and five versions of its website, each tailored to a different audience. The startup declined to disclose how many customers it has.
While Typeface lets a company input details and context so the model can generate more specific and accurate content, there are limits to what it can do. For example, Typeface does not create video content yet, even though more than 80% of businesses use video for marketing, according to a study. Typeface plans to launch generative AI for video and animation creation in the future. Although in its nascent stages of development, a few players such as Movio, Meta and Stable Diffusion have launched generative AI tools for video creation.
During his eight years as CTO at Adobe, and while holding senior leadership positions at Oracle and Microsoft, Parasnis saw companies eagerly integrate breakthroughs in cloud technology and office productivity tools. With DALL-E 2 and ChatGPT going viral last year, he says the time is ripe for companies to incorporate generative AI into their content workflows.
AI for marketing content has traditionally been a crowded industry and Typeface will face more established competitors like Jasper and Anyword. Crystal Huang, a VC from GV who invested in Typeface, says that despite the competition there is “room for multiple billion-dollar companies” in the space.
As generative AI continues to gain traction, more startups using these tools for enterprise applications have launched. Sarah Guo, an early stage venture capitalist and founder of investment firm Conviction, says she has seen at least 30 startups pop up in the space in the last few months. But the true differentiators won’t “glue generative AI onto an existing product thoughtlessly,” she says.
Guo believes that the companies most likely to succeed will rethink the end experience. She says that will happen when the base models like those of OpenAI’s GPT 3.5 and Stable Diffusion’s — which are trained on generic public data and can be used by anyone — are adapted or layered with more specific datasets based on the sector being targeted.
“It’s likely not just going to be a blank text box with an API behind it that isn’t yours,” Guo says.
At Typeface, Parasnis wants to rethink the entire content lifecycle for companies. “We are not just doing a simple wrapper around low level APIs,” Parasnis tells Forbes.
One of the most basic concerns for a company is the safety of its data and its brand image. Businesses want to be sure they are not inadvertently creating content that is inaccurate, plagiaristic or offensive and damaging to their reputation. Parasnis says one of the main reasons why brands are hesitant to directly use OpenAI or Stable Diffusion is because they are open-ended and can generate offensive outputs.
Typeface is built on datasets from Stable Diffusion and OpenAI that contain data that is at risk of copyright infringement. In February, Stability AI was sued by Getty Images for copyright infringement. This might create some reluctance for customer adoption while issues of legal liability are still in limbo. Stability AI has responded to the lawsuit: “Please know that we take these matters seriously. We are reviewing the documents and will respond accordingly.”
“A larger brand wants to have control over what’s being put up there. Is it brand safe? Does it look high-resolution enough? Does it incorporate our colors and brand sentiment? Does it align with existing campaigns? They can’t let it go wild,” says Huang from GV.
To alleviate some of these concerns, Typeface is built on Microsoft’s Azure OpenAI infrastructure, which lets the user set up content moderation controls, data governance and safety checks, a procedure that Typeface conducts for each company separately. Parasnis says Typeface is not responsible for all safety issues because their product and pitch is about giving companies control over their data, training models and the output generated. “Typeface is not the one dictating what every company’s safety rules should be, those companies actually will get a say,” Parasnis says.
This article was updated to reflect that Abhay Parasnis was not the CPO of Adobe; he was only the CTO of Adobe.