Flaviu Radulescu began Runware in 2023 when he was testing a text-to-image firm and realized that, although genAI tech was highly effective, it was gradual in producing pictures. So Radulescu teamed up with Ioana Hreninciuc and launched Runware as a dev device platform that makes a speciality of producing pictures, movies, and audio in actual time.

The corporate has seen a lot progress because it first launched. It has powered greater than 10 billion creations for greater than 200,000 builders, the corporate instructed TechCrunch.

The product lets builders combine Runware’s API into their apps after which generate media belongings via one interface, in order that they don’t should arrange any new infrastructure or keep separate integrations. It has a customized AI inference infrastructure for open-source fashions, and offers day-zero entry (that means as quickly as a mannequin is launched, it may well run on Runware) and aggressive pricing, Hreninciuc, who focuses on operations and GTM, instructed TechCrunch.

On Thursday, the corporate introduced a $50 million Collection A in a spherical led by Daybreak Capital. Daybreak Capital Partner Shamillah Bankiya is becoming a member of the board. Others within the spherical embrace Perception Companions and a16z Speedrun. Runware has raised $66 million in funding to date.

Hreninciuc mentioned the corporate stays aggressive via its pricing, which is “more-cost efficient,” along with having a completely unified API. She mentioned the corporate does this with its Sonic Inference Engine, which runs on customized AI {hardware}. It additionally companions with third-party AI cloud suppliers in order that it may well reroute workloads robotically in case extra reminiscence is required.

“On the software program aspect, we closely optimize mannequin loading and offloading, which lets us assist over 400,000 fashions and make any of them obtainable for inference in actual time,” she continued.

Startups that concentrate on dev instruments for picture and video have been a very sizzling marketplace for VC curiosity these days. Fal.ai, as an illustration, simply raised $140 million at a $4.5 billion valuation, its second big increase in a matter of months. Fal.ai focuses on the breadth of mannequin choices versus customizing for pace. So, Hreninciuc considers her opponents to be Fal.ai and Replicate, a startup that runs open-source fashions in apps with just some strains of code.

These corporations, as Radulescu beforehand instructed TechCrunch, promote primarily based on GPU compute time. Runware as an alternative leans towards the mannequin of Steady Diffusion and Flux, providing a extra cost-per-image generated so that folks will pay for what they want somewhat than shopping for a block of compute time.  

Hreninciuc mentioned the recent capital will likely be used to maintain increasing the corporate’s infrastructure and it hopes to make use of its Sonic Inference Engine to energy over 2 million fashions. The massive objective is to be the API for all AI — in order that any generative AI mannequin can and does run on the platform.

“We’re additionally increasing quickly into new modalities,” she mentioned, including that the corporate goes to increase the present group of round 25 to assist make that occur.

General, Hreninciuc hopes Runware continues to “make it potential for functions to scale to tens of millions of customers whereas really preserving their margins,” she mentioned, including that it helps make the market extra reasonably priced. That “advantages everybody,” she continued. “From the app builders to the top customers, and places highly effective AI into extra individuals’s palms globally.”

This piece was up to date to appropriate the spelling of Radulescu‘s title and to replace what the opponents are.