Summary
The Arduboy creator’s next project uses a neural network trained on footage of a real candle’s flame. After engineering a flexible version of Tetris …
The Arduboy creator’s next project uses a neural network trained on footage of a real candle’s flame.
After engineering a flexible version of Tetris that could be wrapped around a magazine, Kevin Bates’ latest creation is a digital candle that uses a custom AI model to accurately simulate the appearance of a flickering flame across a grid of LEDs. It’s a far cry from the types of devices Bates is typically known for, including the Arduboy, but paying $79 for a credit card-sized handheld that can play hundreds of games is an easier sell than a $149 digital candle that doesn’t quite provide the same ambiance as the real thing, even if it’s much safer.
At the top of the 10-inch tall Digital Candle, which is available now with a black or white circuitboard, you’ll find 144 amber-colored LEDs arranged in a 9 by 16 grid with adjustable brightness. On the other end is a USB-C port for supplying power (a rechargeable battery isn’t included) and an Espressif ESP32-S3 chip that runs an AI model with a tiny neural network trained on footage of a real flickering candle.
The flame simulation never repeats and is unique to each candle since it’s being generated on the fly at all times. It’s not a pre-programmed animation set to loop so the results look very convincing, or as convincing as they can be on a low-res grid of LEDs. The Digital Candle is also open source so you can repurpose it as a clock, a notification screen, or run other simulations like Conway’s Game of Life. That could potentially make its $149 price a little easier to justify if you like to tinker.
Most Popular
This is the title for the native ad
More in Tech
This is the title for the native ad
Top Stories
Source: The Verge
