Akool unleashes enhancements to its AI human 3D avatars connected to LLMs

Akool Inc., a developer of artificial intelligence-driven lifelike human video avatars using cutting-edge technology, today announced enhancements to its interactive capability that allows developers to connect large language models, helping model builders create dynamic conversational experiences.

Founded in 2022, Akool has been at the forefront of creating vivid experiences using lifelike avatars that look and act like humans. This includes showing emotion on their faces, full body movements, and hand and arm motions to convey expression. The company says whole experience is a lot like talking to another person on a Zoom call.

The company provides two types of lifelike avatars. Talking avatars are designed to deliver pre-recorded or scripted messages in a dynamic and engaging manner based on text input or pre-recorded audio. The newly updated streaming avatars provide real-time conversational interaction using AI avatars that can respond directly to questions, which are suitable for active sessions such as customer service or guidance.

In an interview with SiliconANGLE, Akool founder and Chief Executive Jiajun Lu said consumers found static avatars boring and uncanny. His company focused on bringing “humanity” back to customer service by making them interactive and present making them ideal for video creation and engagement.

“A streaming avatar can be created from images or a short video, and together it can be paired with a voice,” Lu explained. “It will also do voice clone and mimic the voice of the actual person and then use that to speak back.”

Under the hood, Akool allows customers to build interactive avatars by uploading documents to customize an LLM based on prompts, should they choose. That can be connected to the company’s in-house large language model. Developers can easily swap that out for other LLM offerings of their choosing, including OpenAI, Lu explained, and the avatar will integrate its responses with the model.

“We have seen lots of adaptations,” Lu said. “The application is pretty wide, so we see companies are using them in customer support, for example, because when you see a person you can interact with it’s easier. I think they just get a better experience. And also with AI agents, they want to put a face on that agent.”

Lu mentioned that aside from customer service, language education was also one of the most commercially successful applications of streaming avatars. People found being able to speak with a lifelike person made it much easier to learn another language. He added that he saw potential for growth in other industries including government and healthcare.

Lu explained that the software is extremely versatile and can be integrated into numerous platforms and devices. For example, they include websites, kiosks, smartwatches and mobile apps.

Akool’s streaming avatars have seen a great deal of commercial success thanks to their low latency, which is a critical factor when dealing with real-time conversation. Humans don’t deal well with long response times, especially for a reply. Lu said the company’s full-body motion capabilities made it stand out from the competition and added that the company planned to enhance this feature in the future.

Image: Akool