Akool Live Camera can translate video calls in real time, swap faces, and get live virtual avatars to mimic human movements

Akool Live Camera uses AI to capture human movement and mimic that movement with a generated virtual avatar in real time.

Akool can also translate speech in real time during a virtual meeting and also provide instant face swapping during a call. The AI technology listens to conversations in one language and instantly translates them into the selected target language, providing real-time, synchronized audio that matches the avatar’s lip movements and facial expressions.

This video generation technology owes its smarts to AI from Akool, a startup based in Palo Alto, California, said Jiajun “Jeff” Lu, CEO of Akool, in an interview with GamesBeat.

“Our main motivation is to enhance the real-time experience and live experiences. For example, you can use avatars to join meetings, you can use video translation to do real-time meeting translations, and lots of other things,” Lu said. “We want to make it so you can’t tell the avatar from the real person.”

The company also offers lip-syncing for avatars in real time, where the avatar lip movements can match the words being spoken by a person in real time, Lu said.

This Akool Live Camera tool is a part of the Akool Live Suite, a first-of-its-kind collection of products that features live, real-time video generation with minimal delay. The suite includes live avatars, live face swap, video translation, and real-time video generation.

“The products we offer are live AI avatars, video translation, face swap and image to video generation, and so on,” Lu said. “We definitely are very competitive in the landscape in terms of human centered videos and things that we do are now available to be in real time.”

It delivers the kind of hyper-realistic visuals you’d expect from OpenAI’s video generation model Sora, but created instantly and in real time, Lu said.

The implications of Akool Live Camera are pretty powerful. For the first time, a sales rep can present in perfect, lip-synced Spanish while speaking only English. A CEO can address global teams as a hyper-realistic digital avatar. A Twitch streamer can broadcast as an anime character without expensive motion-capture gear. And it all happens live in sub-100-millisecond latency across platforms like Zoom, Microsoft Teams, and Google Meet.

“Akool Live Camera sets a new standard in AI-powered video generation technology, going well beyond scripted, text-based prompts,” said Lu. “This opens up a new array of possibilities for virtual meetings and live streams, especially when connecting with international audiences.”

A new paradigm for live AI-powered video generation

Jiajun “Jeff” Lu is CEO of Akool.

Akool Live Camera isn’t merely another video generator. It’s an interactive engine that simulates
human presence dynamically, analyzing live audio/visual inputs to generate responsive avatars with
expressions and contextual awareness.

Akool Live Camera thrives in unscripted environments where minimal latency makes synthetic humans indistinguishable from reality, such as live streams, virtual meetings, and augmented reality gaming. At least that’s the goal, said Lu.

The breakthrough lies in the technology’s ability to synthesize human interactions without preprocessing. Akool Live Camera’s edge-computing architecture processes live feeds instantaneously, allowing the avatars to adjust emotion, gestures, and speech cadence based on real-time audience analytics—a feat akin to an AI director improvising a film during live production.

Key features of Akool Live Camera, all in real-time include:

● AI Avatars: Seamless, photorealistic avatars that mirror a speaker’s expressions, gestures, and tone—reacting dynamically to audience cues in real time.
● Video Translation: Instantly translates spoken language while preserving voice identity and syncing lip movements—enabling lifelike, multilingual communication during live events.
● Live Face Swap: Swaps faces in real time with precision and emotion retention, allowing speakers to represent different identities while maintaining authentic performance. The company worked on applications with Coca-Cola and Qatar Airways.
● AI Video Generation: Creates unscripted, hyper-realistic video on the fly—no pre-recording, scripting, or post-production needed. Content is generated live, based on context, tone, and audience interaction.

Key capabilities of Akool Live Camera include:

● Unmatched live interaction: Live face swap, avatar streaming and multilingual translation during calls/streams outpace other pre-recorded solutions.
● Real-time multilingual translation: Break language barriers with synchronized voice translations that maintain the nuances of your original speech.
● Dynamic expression and gesture mapping: Ensure your avatar reflects your real-time emotions and movements for authentic engagement.
● Cross-platform versatility: Smooth and easy integration with Zoom, Microsoft Teams, Google Meet and more.
● Privacy-forward design: Professional avatars protect user identities in sensitive meetings, with local facial data processing for added security.
● Market- and audience-specific customization: Leverage anime, retro or business-centric avatars with robust outfit/persona swapping.

Lu said Akool Live Camera fundamentally changes the future of live video creation — no longer is it limited to just providing text prompts. The combination of Akool’s AI and intuitive design empowers creators, educators and enterprises to connect more authentically and efficiently than ever.

Slated for general availability in late 2025, Akool Live Camera is set to transform global communication through real-time, AI-powered interactions. Currently in beta and available to a select group of early adopters, the platform offers an exclusive glimpse into the future of live video.

You can secure your early access today at akool.com/live-camera and be among the first to experience the next era of live AI video generation. Secure your early access today by visiting https://akool.com/live-camera.

Origins

Founded in 2022, Akool has grown rapidly and invoiced tens of millions of dollars. Its product lineup includes video translation, real-time streaming avatars, studio-quality face swap, talking avatars, and the newly launched Akool Live Suite—a first-of-its-kind collection of real-time tools enabling live avatars, live face swap, and dynamic video generation with minimal delay.

Unlike Sora, which crafts narratives from text prompts, Akool Live Camera thrives in unscripted environments such as live streams, virtual meetings, and AR gaming. The goal is to take advantage of low latency to make synthetic humans created by Akool become indistinguishable from reality, Lu said.

The company has about 80 people now, with team members who used to work at Apple and Google. Lu himself worked at Google Cloud with a focus on cloud video processing. He also worked at Apple on Face ID. While the headquarters is in Palo Alto, Lu said the team is spread out.

He said the team hasn’t raised much money and is instead generating revenue from AI avatars, face swapping and video translation. Lu said the company can do a wide variety of languages in terms of real-time translation.

“Definitely AI video is moving at a faster pace of change. We are following that pace. In the long run, I believe that having a good user community will be pretty important in the coming years,” he said. “I predict the tech will get mature pretty quickly.”

As a small company, he said the focus is on developing models that are better for the tasks that we people care about.

“We are very ahead in this live game. Definitely, we have very strong engineers [who] are optimizing all the AI to make them run faster. We also have very strong engineers to optimize the whole pipeline to make them work well and have good experiences,” Lu said. “And we build our models from scratch ourselves. From model design to data collection to the whole pipeline, rather than leveraging some open source stuff.”

He said the company checks for copyrights when training models in order to avoid using IP for which it doesn’t have rights.

I asked what Lu thinks about the worries have about AI. He noted AI is getting “high attention” and his goal is to make AI work properly. The company puts watermarks into AI-generated content so it can’t be mistaken for being AI or human. The company also has content moderation tools.

GB Daily

Stay in the know! Get the latest news in your inbox daily

Read our Privacy Policy

Thanks for subscribing. Check out more VB newsletters here.

An error occured.