(Bloomberg) -- Ideogram, a company developing artificial intelligence image generation technology, is raising $80 million from investors — a large round for a startup that’s less than two years old. 

The Toronto-based startup is set to announce its Series A funding round on Wednesday, Ideogram Chief Executive Officer and co-founder Mohammad Norouzi told Bloomberg. Andreessen Horowitz is leading the round, with participation from Index Ventures, Redpoint Ventures, Pear VC and SV Angel. The startup declined to give its valuation. 

Ideogram launched in August 2023, led by several former Google employees who helped create the first version of the tech giant’s Imagen image-generation software. The startup’s initial aim was, in part, to solve a vexing issue with so-called generative AI: letting users create an image with text you can actually read, such as an image of a protester holding a legible sign or a cute cat in a T-shirt that clearly says, “Ask me about my AI startup.”

In September, when its software was publicly released, popular AI image generators such as Midjourney, OpenAI’s DALL-E 2 and Stability AI’s Stable Diffusion would often show nonsense when asked to create pictures including words.

A lot has changed since then. Generative AI — technology that can can produce content like pictures, text and video — continues to develop at a dizzying pace, and some other tools, such as OpenAI’s newest image model, DALL-E 3, can now capably produce text in pictures, too. But Norouzi thinks Ideogram’s latest AI software — which it also rolled out on Wednesday — will keep it ahead of competitors, particularly when it comes to generating images with lengthy, complicated text.The new version of Ideogram’s software shows the correct text “much more frequently than before,” Norouzi said, and it’s also better at producing realistic-looking images.

As part of the new software, Ideogram released a feature called “magic prompt” that can lengthen the written prompts users provide in hopes of producing better and more detailed images. For instance, the feature might elongate a prompt that initially reads, “a photorealistic pika with bumblebee antennae and yellow-and-black stripes” to add several sentences including details like: “The adorable creature stands on its hind legs, leaning slightly forward with a curious expression.” Users can turn the feature on or off, or set it to an “auto” mode that lets the software determine whether or not to trigger a longer prompt.

“We’re trying to be transparent, and we think that’s important,” Norouzi said.Ideogram’s funding announcement comes days after Google paused the ability of its AI system, Gemini, to generate images of people, following criticism that the tool often refrained from creating visuals of White people. As Bloomberg reported Wednesday, Google  tried to reduce bias in part by elaborating on user’s prompts, but did not make it totally clear to Gemini users what was happening behind the curtain.

Ideogram, which also raised $16.5 million in seed funding last year in a round led by Andreessen and Index, plans to put its latest round toward hiring and computing needs. One of the reasons for Silicon Valley’s recent AI funding frenzy is that it’s notoriously expensive to train, run and experiment with cutting-edge AI models. As part of the deal, Andreessen general partner Martin Casado will join Ideogram’s board. 

“The funding enables us to be very data-driven and analytical in really teasing apart what works and what doesn't work,” Norouzi said. He added that the cash infusion will help the company continue offering its AI image-creation software to users for free, though Ideogram has also begun offering paid subscriptions that let users do things like have their images generated more quickly and access editing tools. 

©2024 Bloomberg L.P.