Interactive content startup Flam is looking to build up impressive traction that it says was made while operating in “stealth mode” after announcing a $40 million Series B round of funding today.

The round was led by QED Investors and saw the participation of Claypond Capital, Australian Gulf Capital, the SRK Family Office and angel investors Martin Chavez and Olivier Pomel, the Chief Executive Officer of Datadog Inc. Existing backers including RTP Global and Dovetail also participated in the round.

Flam, which is based in San Francisco, has built a proprietary artificial intelligence infrastructure platform that enables creators to generate interactive videos, real-time three-dimensional experiences and hyper-real visual agents. The platform is able to create these interactive experiences dynamically, in real-time, so that creators can deliver highly engaging content to their audiences at any moment.

Flam founder and Chief Executive Shourya Agarwal said he built the company after making the simple observation that digital content has evolved in waves over the years. First we saw simple text, then we saw images, and finally video also became a mainstay of the internet. But each wave led to the creation of media that only flows in one direction – from creators to viewers. Despite this, Agarwal believes there has always been demand for more interactive content, but the problem is that creators haven’t had any way to do this.

This is what Flam is changing. It has developed an infrastructure stack backed by more than 15 patents that allows creators to build programmatic content experiences across any digital medium. The startup calls these experiences “flams,” and says there are three distinct types.

The first is Flicks, which are interactive videos generated by AI models, with AI compression algorithms that optimize playback, ensuring that interactions remain instantaneous and smooth. The second type of flam is Airboards, which refers to high-fidelity 3D content experiences that can be streamed directly to user’s devices and their camera interfaces. Flam explained that these assets can be generated via text and image prompts using its proprietary model Fable.

Last, Flam also offers Visual Agents, which are described as “hyper-realistic avatars” that can engage with users in real-time conversations, similar to a video call. The Visual Agents are powered by an identity-preservation and motion-transfer model called Fantom, which generatively streams facial expressions, the company said. Fantom works in conjunction with a 26 billion-parameter mixture-of-experts model called Falcon, which is capable of blazing-fast 30 millisecond time-to-first-token inference speeds, enabling those Visual Agents to take actions in real time.

Agarwal said technology has evolved far enough to support people’s natural curiosity. “We don’t just want to see the world; we want to reach out and shape it,” he said. “Passive media was a limitation of technology, never of desire. We’re doubling down on the AI research that makes interactivity feel effortless, and bringing Flam to every enterprise that wants its content to invite people in, rather than talk at them.”

Though Flam says it has officially been operating in stealth mode until now, it has generated impressive traction that belies that statement. In the last year-and-a-half, it has signed up more than 100 enterprise customers, including Google LLC, and claims to be fast approaching $100 million in annually recurring revenue, a target it hopes to reach next year.

The funds from today’s round will help to support ongoing research and development, expand Flam’s product suite and create new “flams,” and expand its sales operations. Ultimately, Flam hopes this round will help it to build the foundational infrastructure layer for every brand to create and share interactive content.

QED Investors Managing Partner Nigel Morris said many startups have tried to deliver truly interactive content experiences in the past and failed. “The pattern has always been the same: compelling demos that never scale,” he said. “What convinced us was the combination of a differentiated AI infrastructure, a platform enterprises can deploy without an engineering lift, and the commercial traction to prove it.”

Images: Flam