The Silicon Valley lab teaching machines to subtitle, dub and interpret the way humans actually speak - in 45+ languages, often before the sentence even ends.
Somewhere right now a conference is running in three languages at once, a FAST channel is pushing a Korean drama into a dozen markets, and a Zoom webinar is being read by people who do not share a single word of the speaker's vocabulary. The connective tissue, increasingly, is XL8.ai.
Most translation software treats language like a spreadsheet: feed it a cell, get back another cell. XL8 started from a more inconvenient truth - people on screen do not talk in spreadsheets. They interrupt. They use slang. They trail off. They say the opposite of what they mean and expect you to laugh. A machine that handles paperwork well will butcher a sitcom.
So when Tim Jung left Google in 2019 and teamed up with former Apple engineer Jay Park, they made a deliberately narrow bet. Not general-purpose translation for everything. Translation built specifically for media, broadcast and live events - the messy, time-coded, context-soaked stuff that big models tend to flatten.
The contrarian move sits in the training data. While much of the industry races to scrape ever more of the open web, XL8 trains its engines on what it calls "golden" data: 100% human professional-curated material. The wager is simple and a little old-fashioned - that quality beats quantity when the output has to sound like a person, not a press release.
Two products carry the thesis. MediaCAT turns a raw video into finished, subtitled or dubbed content through three moves it labels plainly: sync, translate, dub. EventCAT does the live version, beginning to translate word-by-word before a speaker finishes a thought, then quietly correcting itself for context as the sentence lands.
"Our goal is to help localization professionals collaborate quickly and easily." Tim Jung, Co-Founder & CEO, XL8.ai
XL8 splits the world into recorded and live. Each gets a purpose-built engine - and both refuse to sound robotic.
An AI localization workflow built on three steps: speech-to-text transcription (Sync), machine translation (Translate) and synthesized voice (Dub). Upload a file, get back subtitles or dubbing with automated quality checks. "Projects 2.0" adds collaborative, generative-AI-powered project management for whole localization teams.
Real-time interpretation and live subtitles in 50+ languages. A low-latency engine translates before sentences finish, while staying context-aware. EventCAT Conference handles big in-person events; EventCAT Online Meeting plugs into Zoom, Google Meet and Microsoft Teams - including a dedicated Zoom app.
Domain-specific MT engines trained on a customer's own curated data, plus API integration and on-premise options for studios that cannot let pre-release content leave the building. Backed by SOC 2 Type II and TPN Gold Shield security.
Both founders hold Computer Science degrees from Columbia University - then split to opposite ends of Big Tech before reuniting around one problem.
Mission: create world-class AI for localization, broadcast and live events, making communication accessible across every language and region.
Vision: eliminate language barriers through AI to help the world communicate. The team spans Silicon Valley and an international engineering office opened in Seoul in 2021.
No moonshot mega-round - a seed, a bridge, and a Series A led by Korea's KB Investment, with Atinum returning at every step. Earmarked for generative AI and expansion into Europe, the Middle East and the Americas.
INVESTORS: KB INVESTMENT (LEAD, SERIES A) · ATINUM INVESTMENT · MICROSOFT FOR STARTUPS FOUNDERS HUB
XL8 Inc. founded in Silicon Valley by Tim Jung and Jay Park.
International engineering office opens in Seoul, Korea.
$3M pre-Series A bridge round to accelerate machine translation in media localization.
Unveils "Projects," the next-gen MediaCAT platform powered by generative AI.
Closes $7.5M Series A led by KB Investment - total funding reaches $11.5M.
MediaCAT wins the AI & Data category at the Products That Count Product Awards.
EventCAT showcased at World IT Show 2025 as a real-time multilingual interpretation platform.
EventCAT 2.0 officially launches its AI real-time interpretation solution.
Return to the room running in three languages at once. A decade ago the audience would have worn headsets, waited on tired human interpreters, and missed every joke by half a beat. Today the subtitles arrive as the speaker talks, the dubbed track sounds like a person who meant it, and the Korean drama reaches a dozen markets before lunch. XL8 did not invent the desire to be understood across borders - it just made the delay disappear. Twenty-two people, $11.5 million, and a quiet insistence that "good enough" translation is not actually good enough. The sentence finishes. So, finally, does the wait.