Flaviu Radulesc started Runware in 2023 when he was testing a text-to-image company and realized that, though genAI tech was powerful, it was slow in generating images. So Radulesc teamed up with Ioana Hreninciuc and launched Runware as a dev tool platform that specializes in generating images, videos, and audio in real time.

The company has seen much growth since it first launched. It has powered more than 5 billion creations for more than 100,000 developers, the company told TechCrunch.

The product lets developers integrate Runware’s API into their apps and then generate media assets through one interface, so they don’t have to set up any new infrastructure or maintain separate integrations. It has a custom AI inference infrastructure for open-source models, and provides day-zero access (meaning as soon as a model is released, it can run on Runware) and competitive pricing, Hreninciuc, who focuses on operations and GTM, told TechCrunch.

On Thursday, the company announced a $50 million Series A in a round led by Dawn Capital. Dawn Capital Partner Shamillah Bankiya is joining the board. Others in the round include Insight Partners and a16z Speedrun. Runware has raised $66 million in funding to date.

Hreninciuc said the company remains competitive through its pricing, which is “more-cost effective,” in addition to having a fully unified API. She said the company does this with its Sonic Inference Engine, which runs on custom AI hardware. It also partners with third-party AI cloud providers so that it can reroute workloads automatically in case more memory is needed.

“On the software side, we heavily optimize model loading and offloading, which lets us support over 400,000 models and make any of them available for inference in real time,” she continued.

Startups that focus on dev tools for image and video have been a particularly hot market for VC interest lately. Fal.ai, for instance, just raised $140 million at a $4.5 billion valuation, its second giant raise in a matter of months. Fal.ai focuses on the breadth of model offerings versus customizing for speed. So, Hreninciuc considers her competitors to be general open-source model makers, such as Hugging Face; Replicate, a startup that runs open-source models in apps with just a few lines of code; and Together AI, a similar genAI model hoster.

These companies, as Radulescu previously told TechCrunch, sell based on GPU compute time. Runware instead leans toward the model of Stable Diffusion and Flux, offering a more cost-per-image generated so that people can pay for what they need rather than buying a block of compute time.  

Hreninciuc said the fresh capital will be used to keep expanding the company’s infrastructure and it hopes to use its Sonic Inference Engine to power over 2 million models. The big goal is to be the API for all AI — so that any generative AI model can and does run on the platform.

“We’re also expanding rapidly into new modalities,” she said, adding that the company is going to expand the current team of around 25 to help make that happen.

Overall, Hreninciuc hopes Runware continues to “make it possible for applications to scale to millions of users while actually keeping their margins,” she said, adding that it helps make the market more affordable. That “benefits everyone,” she continued. “From the app builders to the end users, and puts powerful AI into more people’s hands globally.”