Beeble AI bags $4.75M to bring Hollywood VFX to iPhones

Hollywood shot The Mandalorian inside a giant, multi-million-dollar LED cylinder called the Volume. Actors stepped on set, virtual worlds wrapped around them, and realistic ambient lighting spilled across their faces in real time. It looked phenomenal, but unless you have an eight-figure Disney budget, you are stuck with green screens and bad spill light.

A South Korean VFX startup called Beeble AI just raised $4.75 million in seed funding to bring that caliber of virtual production down to anyone with an iPhone. The round was led by Basis Set Ventures, with participation from Fika Ventures, valuing the tiny seven-person team at $25 million.

Beeble was founded in 2022 by five former AI researchers from Krafton, the gaming giant behind PUBG. While everyone else in the AI video world rushed to build prompt-to-video generators, this team spotted an overlooked problem that drives cinematographers crazy: lighting.

Inverting the LED wall

Traditional virtual production brings the digital environment into the real world via massive LED walls. Beeble flips the equation completely: it pulls real-world actors directly into the virtual environment.

At the heart of the platform is an AI foundation model that the team showcased at CVPR 2024. The model takes ordinary 2D footage of an actor shot on a regular phone and reconstructs it into a physically accurate 3D representation complete with precise geometry and surface textures.

Once the actor is digitized, filmmakers can alter light sources, reframe camera angles, and swap digital backdrops directly in post-production. The desktop tool, currently called SwitchLight Studio, is set to be rebranded as Virtual Studio in the third quarter.

Their consumer photo app, SwitchLight, has already clocked roughly 3 million downloads and started pulling in revenue last October. Even third-party tools like Caption AI are integrating Beeble's relighting SDK.

Why prompts cannot fix film

Beeble CEO Hoon Kim did not mince words when discussing text-to-video engines like OpenAI's Sora or Runway. He pointed out that while text prompts are great for parlor tricks, they are infuriating for serious directors.

Generative AI video models are notoriously random. Ask them to tweak a subtle shadow, and they might regenerate the actor's face entirely. Kim argues that compelling storytelling requires surgical control over every single variable: the exact position of a rim light, the focal length of the virtual lens, and the continuity of the subject from shot to shot.

Instead of catering solely to indie film sets, Beeble is also targeting micro-drama streaming platforms like ReelShort, where churn is fast, budgets are razor-thin, and turnaround times leave zero room for extensive physical lighting rigs.

Why it matters

The generative video space is drowning in hallucinated eye candy, but actual working filmmakers cannot afford unpredictability. Beeble's move is clever because it targets the most expensive friction point in indie filmmaking—lighting and staging—without trying to replace the entire creative workflow with a text box.

If you are an indie director or a solo YouTube creator, this could realistically eliminate the need for costly studio rentals. Being able to shoot an actor in your garage and accurately relight them to match a sunset or a cyberpunk alleyway changes the budget math overnight.

That said, we think the platform faces stiff reality checks. Turning 2D video into flawless 3D geometry is notoriously fragile when handling fast motion, wispy hair, or reflective fabrics. On top of that, established VFX software suites are rapidly shipping their own neural relighting tools. For a seven-person startup, maintaining an algorithmic lead against deep-pocketed 3D incumbents will be a brutal sprint.