Notes from building real-time AI avatars at Anam: low-latency systems, product engineering and the distance between a demo and something people can use.
By Harry Smaje, product engineer at Anam.
Four leopards in a day, wild dogs, lions on the road in the early morning, an hour with a cheetah mother and her three cubs, and a serval in the spotlight. A day-by-day diary of a week at Kruger National Park.
What changed in Anam's new avatar model, and what it took to turn those changes into a product people can actually use.
What Anam's Cara 3 avatar model does, how it turns audio into a speaking face, and where model generation fits into a live conversation.
Patterns and mistakes from two years of helping companies build with AI.
What it's actually like building an AI product at a startup where everyone does everything.
Our testing at Anam pointed to a 200–300ms sweet spot for conversational timing. Here's why that leaves so little room for the engineering.
What we learned from personalised customer demos, an Anam expert and a guessing game on our homepage.
I'm joining Anam to build out the Lab product and SDK experience, bringing my background as an engineer and founder to real-time AI.
Five days exploring Singapore, with Universal Studios, Chinatown, Little India and plenty of food stops, followed by Ubud, sunrise on Mount Batur and the beach in Nusa Dua.
Our Florida holiday together: Guardians of the Galaxy, ten goes on Everest, matching shoes with Sulley and plenty of time around the four parks.
Prefer a feed? Subscribe via RSS