How Deeptrack’s Suite Of Devices Tackles A Rising Deepfake Disadvantage

Deeptrack is solely ten months earlier. It was born out of a layoff, a $10,000 cheque, and a wager on a rising disadvantage: deepfakes. Within the current day, the Nairobi-based startup helps buyers affirm media, detect forgeries, and stop synthetic fraud sooner than it spreads. Its founder, Brian Okay’Oyundi, is conscious of the job isn’t close to executed.

“We’re 10 months earlier proper now,” he instructed me all through an interview closing Friday. “One of the best half is how Deeptrack was born, from me getting laid off closing 12 months from Vambo AI,” a Kenyan AI startup. “I really feel I started setting up Deeptrack two weeks later. Inside that interval of getting laid off, which was mid-October, I landed an individual investor who cut back me a check of $10,000, and I started setting up.”

The worldwide deepfake detection market is projected to develop from $114.3 million in 2024 to over $5.6 billion by 2034 as organisations demand devices that affirm authenticity in digital content material materials. North America leads the market, accounting for 42.6% of earnings. Deepfake detection devices are designed to establish manipulated or utterly faux audiovisual content material materials created using AI, now a rising menace in media, politics, and cybersecurity.

The startup wrote its first traces of code in November 2024 and its site went dwell within the equivalent month. Since then, it has launched 4 merchandise: Sentinel, which helps with know-your-customer (KYC) checks; Gotham, a deepfake detection system constructed for enterprise use; Atlas, a instrument for verifying the authenticity of media recordsdata; and Foundry, which flags forged or altered paperwork submitted in insurance coverage protection claims.

Deeptrack presently has two paying buyers preserving the lights on. The group is lean, about ten people, unfold all through numerous African nations, along with Burundi, Morocco, and Cameroon. 

How Deeptrack works

Deeptrack’s detection system runs on deep neural networks educated to establish indicators that individuals usually miss. It makes use of a multistage forensic pipeline, starting with pre-processing to standardise inputs. This consists of resizing, format conversion, physique extraction for video, and fingerprinting for traceability.

Subsequent is the detection half, the place Deeptrack’s fashions, educated on tons of of deepfake samples, run alongside APIs from Actuality Defender, which scan for indicators of manipulation. Looking for these APIs is a smart choice for a small group because of it saves time, lowers costs, and frees them as a lot as assemble devices for use circumstances like KYC and political deepfakes.

The final word steps include attribute analysis and classification. The system checks for indicators like irregular eye movement, harmful lighting, or lip sync factors. Each file is then scored using Deeptrack’s perception model, serving to buyers resolve whether or not or to not behave, flag, or archive the content material materials.

Gotham and the hazard of enterprise abuse

One among Deeptrack’s merchandise, Gotham, is constructed for enterprises, significantly media properties, security corporations, and banks. Gotham, by the use of an internet app, permits bulk uploads and multi-modal detection, which implies it might properly course of every video and audio on the same time.

One use case is media vetting, the place, for example, a shopper uploads a video or audio clip to substantiate if it has been tampered with or synthetically generated. Deeptrack’s model runs an analysis and sends once more a finish outcome with a confidence score and tampering report.

That rigidity between detection and abuse, the place deepfakes are used to harass or discredit others, is one factor the startup thinks about. In a number of circumstances, deepfakes are exhausting to detect and as they get additional convincing, detection devices wish to improve merely as fast to take care of up.  

Get the perfect African tech newsletters in your inbox

Why starting small helps

When Okay’Oyundi launched Deeptrack, he made a reputation that didn’t age properly: launching with an infinite group, which included a bunch of acquaintances that he wanted to let go in early 2025. “We hadn’t constructed one thing clients wanted however, nevertheless we’ve been already spending on salaries and operations.” By December 2024, that that they had burned by the use of half the money they’d raised.

The group, which contains advisory members, has since grown slowly, with most of them working remotely and their roles now narrowly outlined. They’ve moreover turn into additional frugal, consistent with Okay’Oyundi, and are however to spice up any additional funding to date.

Mirror and previous

Deeptrack’s subsequent launch is Mirror, a instrument aimed towards high-net-worth folks, public figures, and firms coping with standing assaults. The instrument scans for faux audio or video impersonations circulating on-line and flags threats sooner than they go viral.

Mirror operates on two concepts. First, proactive scraping shows social media channels or chat apps that always share manipulated content material materials. Second, voiceprint and faceprint security, which incorporates making a biometric fingerprint of enrolled clients and flagging any media that makes an try to imitate their voice or facial choices.

What comes subsequent

The group is making able to scale by hiring additional machine finding out engineers and growing its office home. Deeptrack may additionally have to find an even bigger, secure home to take care of its servers. 

“Correct now, we work out of a small home in Nairobi,” Okay’Oyundi acknowledged, referring to a co-working home the place I met him alongside along with his group. 

Okay’Oyundi has moreover been accepted into The Residency, a six-week in-person founder programme in San Francisco, and plans to utilize the time to be taught to pitch Deeptrack for follow-on funding. “We’re occupied with elevating later this 12 months. One factor small, probably $300,000. Merely adequate to extend the runway and do additional hires.”

Mark your calendars! Moonshot by TechCabal is once more in Lagos on October 15–16! Be a part of Africa’s excessive founders, creatives & tech leaders for 2 days of keynotes, mixers & future-forward ideas. Early fowl tickets now 20% off—don’t snooze! moonshot.techcabal.com