A webcam points at a person. On the monitor, a fox raises an eyebrow at exactly the same moment. The trick looks lighthearted, almost disposable. Behind it sits a stubborn collection of hard problems: face tracking, rendering, rigging, timing, model formats and the curious human ability to notice when a smile arrives half a beat late. Dragos-Florin Stanculescu has spent more than a decade inside that gap between a moving character and a character that feels occupied.
His route there began in Romania’s young game industry. In 2001, after studying engineering, he joined FUN Labs. Over 13 years he worked through concept art, textures, visual effects, technical art and direction. The titles changed, but the useful skill stayed consistent: translating what an artist wants into what software can produce in motion.
He also kept tinkering with faces. Game characters already had animation systems, while professional motion-capture tools sat far beyond the reach of an ordinary computer owner. Stanculescu and a group of colleagues began asking whether those pieces could be brought home. A regular camera could supply the signal. A virtual camera could pass the result into chat and streaming software. A digital creature could do the performing.
Five colleagues and a camera
The first FaceRig core was five people with long experience in games: two artists, two programmers and Stanculescu as technical artist. Their shared history mattered. They did not need to discover one another’s working habits while discovering the product. They built during off-hours, then moved toward full-time work without abruptly abandoning the teams that employed them. It was a cautious start for a product with a theatrical premise.
Crowdfunding supplied the first public test. The campaign asked for $120,000 and finished above $300,000. Backers were buying more than a novelty. FaceRig placed a compact performance studio behind a consumer webcam. Your facial movement drove a 2D or 3D model, and the resulting virtual camera could travel into the software people already used.
By 2016, Stanculescu said FaceRig had generated more than $2 million in sales. By 2018, Holotech reported four million combined downloads across desktop and mobile. That year, a $2 million Series A led by SignalFire gave the Bucharest company money to expand the product and its presence in San Francisco.
The numbers describe adoption. They do not explain the appeal. FaceRig let the operator borrow a shape without surrendering the signals that make conversation feel live. It could be funny, useful for a stream, or simply a different way to enter a call. Stanculescu’s enduring question was whether the expression could survive the disguise.
“We don’t want to be a walled garden, we want to be an open standard.”Dragos-Florin Stanculescu on Animaze
The expression has to make it through
Stanculescu’s avatar philosophy is grounded in performance. A pre-baked grin triggered by a button may be legible, but it does not carry the same information as a smile forming on a face. The eyebrow, pause and head tilt are not decoration. They are part of the sentence. A useful digital identity has to transport those details while giving its operator control over appearance.
One inspiration arrived from outside the production pipeline. A 2010 talk by game designer Jane McGonigal explored the intensity and confidence people can bring to games. Stanculescu focused on the feeling of stepping into a character and then carrying some of that character’s agency back out. He had grown up interested in coding, games and digital art. The avatar became a bridge between those crafts and a broader question about how people present themselves.
He also connects avatars with privacy. Growing up in communist Romania gave him a concrete reference for the risks attached to recognized speech. A digital mask can separate an idea from a physical appearance. In his formulation, that separation need not flatten the speaker. The face can be hidden while timing, emotion and intent remain available to the listener.
Can a user choose the surface of a character while the software preserves the texture of the performance?
Why the first success needed a rebuild
FaceRig proved there was demand, then revealed the economics of serving it. A $15 desktop product made advanced-looking animation approachable. It also limited how much Holotech could spend on the harder layers. Better tracking requires specialized engineering. Convincing models require artists and riggers. Every extra controlled bone, blendshape and import path adds labor before it adds delight.
Holotech rebuilt the premise as Animaze and released it in late 2020. The new product kept a free tier while using subscriptions for deeper creator and commercial needs. Stanculescu’s defense of the model was direct: the company needed to pay programmers and artists, and it did not want an advertising business that depended on peering into user data.
The shift frustrated some FaceRig users who preferred a one-time price. It also made the trade visible. Cheap software can spread a category while starving the next technical step. A subscription can fund ongoing work while raising the burden to prove value every month. Holotech chose to keep the doorway free and charge users who needed longer virtual-camera sessions or professional features.
“Pixar in real time.”The compact phrase Stanculescu uses for Holotech’s animation ambition
That four-word target is useful because it is obviously unfinished. Pixar’s work is composed, reviewed and rendered. A live avatar gets one chance. It has to read a moving face, interpret incomplete signals, solve a rig and render the result quickly enough for conversation. It must also support models made elsewhere, from Live2D characters to 3D formats and custom pipelines.
Owning its animation and rendering layer gives Holotech control over that pursuit. Stanculescu has described Animaze as a broad interaction engine rather than a sealed collection of characters. The company can support external model creation, different visual formats and new tracking inputs without treating a general-purpose engine’s roadmap as its own.
The character comes from somewhere else
Avatar software sits in an ecosystem of artists, modelers, streamers and toolmakers. Holotech could supply every character itself, but that would narrow the range to one studio’s output. By 2016, the pull of outside creation was already visible. At a Live2D event, Stanculescu said that 150 of the 170 avatars then available through FaceRig’s Steam Workshop used Live2D. The community had selected a practical format for bringing illustrated characters to life, and FaceRig had become one of the places where those characters could perform.
Animaze carried that lesson forward with import paths for 2D and 3D work, support for custom rigs and a marketplace alongside Holotech’s own catalog. The user-facing promise sounds simple: bring the character you want and make it move. The production reality spreads across drawing, mesh construction, bones, blendshapes, physics, lighting, tracking calibration and usage rights. Each format arrives with a community and a set of expectations. Openness here is less a slogan than a continuing compatibility bill.
The surrounding stack also includes sound. A 2022 partnership brought Voicemod filters and sound effects into Animaze, pairing a chosen face with a changed voice. The collaboration fit Stanculescu’s description of avatar identity as both seen and heard. It also showed how Holotech could extend the experience through connections instead of building every component internally. A rendering engine, a model ecosystem, a tracking device and an audio tool can come from different makers. The performance only feels unified if their seams stay out of the conversation.
A product still in motion
The 2026 Animaze roadmap still reads like the work of people living with the machinery. One planned project is a lighter iOS app focused on tracking, stripped of the features that make sense when the phone is the full avatar device but become clutter when it is only a camera for a Windows PC. Other updates have improved motion filtering, export codecs, account security and resilience when network services become unreliable.
The release notes are signed by a short list of first names: Alex, Catalin and Dragos. New original avatars arrive beside infrastructure fixes. A physics-driven hat may share a season with AV1 export and connection resilience. That mixture captures Holotech’s business. The character is what the user meets. The engineering is what allows the meeting to keep happening.
Stanculescu’s career now spans engineering education, concept art, effects, technical direction, fundraising, pricing arguments and product leadership. The common material is translation. At FUN Labs, it meant turning art into a game-ready asset. In FaceRig, it meant turning webcam data into a character. In Animaze, the target is more demanding: turning a person’s chosen appearance into a live identity without losing the person’s cadence along the way.
The fox on the screen raises its eyebrow. The motion is tiny. It is also the whole point. A model can be purchased, imported or drawn. A mask can make its wearer unrecognizable. The difficult work is making the next person in the call understand what the wearer meant.