I created an interactive digital avatar of myself — and you can talk to it

Interactive AI avatars can now be created from individuals, offering a way to automate personalized content delivery.

Video generation startup Synthesia has created an interactive digital avatar of TechCrunch reporter Dominic-Madori Davis. This avatar is trained on a specific article Davis wrote about venture-backed startups and fraud. Users can ask the avatar questions about the article, and it will respond based on its training data. The process involved a mini film studio session where photos and a two-minute voice recording were captured. Synthesia offers various avatar products, including static avatars that read scripts and interactive avatars powered by voice-to-text, language, and text-to-voice models. The company, valued at $4 billion and with over $100 million in ARR, also allows enterprises to integrate these avatars into training and role-playing scenarios. Synthesia's technology stack can incorporate third-party models from companies like OpenAI and ElevenLabs, and clients can choose their cloud hosting options.

The development of interactive digital avatars, exemplified by Synthesia's creation of Dominic-Madori Davis's avatar, signifies a significant leap in applying AI to PR and content creation. This technology moves beyond static AI-generated text or simple scripted avatars to create dynamic digital representations capable of engaging in conversations based on specific training data. Synthesia's offering, which integrates voice-to-text, language processing, and text-to-voice models, allows for a degree of personalization that blurs the lines between human and digital interaction.

This advancement suggests a future where personalized AI assistants or representatives could become commonplace across various industries, from customer service to journalism and sales training. While the current iteration is deterministic, meaning it sticks to its training data, the potential for more fluid, non-deterministic avatars powered by advanced language models raises questions about authenticity, trust, and the future of human roles in communication and information dissemination.