Neural Garage (VisualDub): Synchronizing Sight and Sound in Post-Production

In this interview, Mandar and Subho, Co-Founders of Visual Dub, unpack how their AI technology is redefining the dubbing experience by eliminating “visual discord”—the jarring mismatch between audio and facial movements in dubbed content. Born out of a pandemic-era insight and powered by deep expertise in facial recognition, Visual Dub aligns lip sync and performance across languages, enabling actors to appear natively fluent in any tongue. From working with top brands to prepping for international expansion, they share their journey from startup to TechCrunch Disrupt stage. This is the written version of the Vitrina podcast interview with the founders of Visual Dub—summarized for quick reading and reformatted into a Q&A style for clarity.

Mandar:
“Visual Dub is solving a longstanding problem in dubbed content—what we call ‘visual discord.’ We’ve all experienced the disjointedness when watching content like a Korean film dubbed in English—the audio doesn’t match the facial expressions or lip movements. Our solution fixes that. Using AI, we ensure that when you watch a dubbed film, it appears as if it was shot originally in that language, creating a seamless experience.”

Subho:
“The idea is to expand content beyond its geographic and temporal limitations. By aligning speech and visual expressions, we let creators go global with a click. There are no Hollywood or Bollywood stars—everyone becomes a global star with this technology.”

Mandar:
“This is a massive opportunity. Take the example of the Korean film Parasite. Its director, while accepting the Oscar, pointed out that subtitles are the biggest barrier to global content consumption. Netflix’s Ted Sarandos said something similar—content should not feel ‘dubby.’ Our tech can turn a dubbed Tom Cruise movie into what feels like a film originally shot in Japanese. This affects theatrical releases, streaming, broadcast trade—basically the whole long tail of global entertainment.”

Mandar:
“If Money Heist Season 10 is shot in Spanish and distributed in 100 languages, it no longer remains a Spanish show. It becomes a show that feels native in every single one of those languages.”

Mandar:
“Our tech creates a new category in post-production. We sit between dubbing and distribution. Once content is dubbed, Visual Dub takes over to sync video and audio, making it look native in each language. There’s also a use case in dialogue correction—reshoots are expensive. Our tech avoids that by fixing performance and dialogue alignment in post.”

Subho:
“We’re separating performance from telegraphy. This unlocks casting opportunities for actors who don’t speak a particular language. Our software helps preserve an actor’s original performance while making corrections or replacements in post-production.”

Mandar:
“During COVID, content pipelines were dry, and one of our co-founders, Anjan, binged a lot of dubbed Korean content. That’s when it hit us—what if this visual discord problem could be solved? The founding team had experience in facial recognition tech, so we began development in late 2020.”

Subho:
“The challenge was enormous. We’re targeting high-resolution screens like TVs and theaters—so any visual artifact would be glaring. We needed our models to preserve facial fidelity—dimples, smile lines, everything. Our algorithm was trained to honor an actor’s unique features while manipulating facial movements at the phoneme level.”

Subho:
“We don’t just work at the language level—we go deeper, down to phonemes, the smallest unit of sound. Every language is a mix of sounds, and those sounds dictate lip and facial movements. So whether it’s 15 languages or more, our input is one video and one audio at a time, and the output is a version of that video that looks visually native in that language. It even adapts if a character shouts in one language and whispers in another.”

Subho:
“In the beginning, we only focused on lip sync. But we realized there was massive value in using AI to match the actor’s original voice identity to the dubbed performance. So while we retain the dubbing artist’s emotions and delivery, the final audio is in the original actor’s voice—say, Tom Cruise speaking Arabic in his own voice. It completely elevates realism.”

Subho:
“One standout example was with Coca-Cola and Harsha Bhogle. They created contextual, post-match video messages using the same footage but different messages. Our tech helped them repurpose it rapidly. Another was with UltraTech Cement and Shah Rukh Khan—the ad was shot in Hindi and recreated in English with his approval. It’s been airing regularly, and that’s a big validation.”

Mandar:
“Right now, we offer a tool to a small group of premium clients. It’s like a video editor—input a video and audio, and get back a synced output. Our North Star is to offer it as a downloadable software, licensed like Adobe Premiere or DaVinci. This will give post-production houses a plug-and-play solution right in their existing workflows.”

Mandar:
“We started targeting film and streaming, but ad agencies were our first adopters. India’s advertising is mostly shot in Hindi and dubbed into 10–12 languages. Our tech solved the regional disconnect. Brands with high standards tested us and were willing to pay. Studios like Excel Entertainment and Prime Focus have also begun testing our tech for upcoming major films.”

Subho:
“We built our algorithms to handle varied angles, lighting, and head turns. Training data was key. Now, we’re at a point where even side angles work without visible glitches. Our goal is always 100% visual consistency—even the actor’s dimples must remain. If Shah Rukh Khan approves an ad, that’s the bar.”

Mandar:
“Our next step is opening an office in LA, the entertainment capital. We’ve already had interest from dubbing and VFX studios—some even explored partnership or acquisition. We’ll follow the post-production supply chain for integrations. Japan, Korea, and the Middle East are also key markets, and we’ll participate in events like MIPCOM and MIPTV.”

Mandar:
“I didn’t fully grasp the importance at first, but Subho was thrilled. It was a major platform.”

Subho:
“TechCrunch Startup Battlefield has been the launchpad for giants like Dropbox and Intuit. We were the only video and entertainment startup selected this year and one of only three from India. The exposure, networking, and validation from being part of such an ecosystem in San Francisco was phenomenal.”