Denis Shiryaev is a Gdansk-based Russian entrepreneur who founded neural.love to clean up vintage film footage using open source software and AI models, including the splendidly-named DeOldify. Since their work started to become recognised, neural.love has restored old footage of Amsterdam and Moscow and more recently they helped PBS in the U.S. with a documentary about women’s suffrage.
The neural.love workflow has recently been speeded up significantly. By upgrading to the latest NVIDIA RTX 6000 GPU they can now process 60 minutes of video in less time than it took an older GPU to do 90 seconds. The same hardware is also used to train their AI model; they can now do in eight hours what used to take a week.
neural.love’s future appears to lie in a cloud-based upscaling service that anybody can use to enhance old VHS tapes or home movies. They hope to be up and running in a few months.
Source: NVIDIA blog