The face on the screen looks real, feels real, and responds in real-time. Its voice is warm, its eyes tracking naturally, and its expressions are shifting ever so slightly as it speaks. This is Bengaluru-based Gnani.ai’s new self-cloned digital human, an AI-generated avatar that delivers expressive, lifelike conversations with micro-expressions, lip sync, and gaze control.
Unveiled at the Global Fintech Fest, the avatar is built on the company’s proprietary platform HumanOS, using a source video and a short voice reference, which allows it to replicate a person’s look, tone, and style of speaking. The avatar is interactive, speaks in 14 different Indian languages, and can be adapted to suit various enterprise requirements.
“Most automated interactions used to be text, then it evolved to voice, and now we’re adding a video layer to make them more real,” said co-founder and CEO Ganesh Gopalan.
While traditional chatbots or voice assistants can make the lack of human touch more evident, Gnani’s AI avatar not only replicates voice and accent, but it also captures look, feel, and expressiveness, making AI-driven conversations much more natural.
Enterprise Impact
The application of such avatars is wide. For sectors like banking, telecom, and retail, it can handle use cases such as video KYC, onboarding, and customer support, while retaining the warmth of a human interaction.
Capturing Micro-Expressions and Tone
The underlying technology is a system that captures a few key facial motion patterns, such as how vowels and consonants shape lip and jaw movement of the person, from a short video of the real person speaking. It captures those moments and mimics the pattern for the rest of the speech.
Typically, there are only three or four ways our facial expressions change while speaking, one of its engineers explained. “For each avatar, we also capture 12 unique vocal parameters, such as tonality, pitch, frequency, and emotional variation,” he added.
Backed by $7.7 million in funding, Gnani.ai was recently selected under the IndiaAI Mission to build sovereign large language models. The company has been working on a multilingual, voice-first, emotionally intelligent AI model, focusing on Indic languages.
