Tag Archives: neural network
Artificial Intelligence Creates an Infinite Bass Solo
The pinnacle of human achievement: we’ve finally reached it. What we have here is a never-ending bass solo generated by a recurrent neural network (RNN). It was trained on two hours of bass improvisation by musician Adam Neely. I just listened to four hours straight and my face is officially melted. Free Bird!
Developed by Youtubers DADABOTS (aka CJ Carr and Zack Zukowski), the live-stream provides an infinite fast-paced bass solo that’s sure to impress your bandmates when you pretend it’s actually something you recorded.
Obviously, I just started a jam band and am using this artificial intelligence generated bass as the backbone for all my songs. I just stand on stage and pretend to rock out while the technology does all the work for me. You know, because why do any heavy lifting if you don’t have to?
[via Engadget]
An AI is livestreaming a never-ending bass solo on YouTube
AI gauges the mental health of cancer patients through eye movements
AI can detect COVID-19 by listening to your coughs
Facebook uses AI to forecast COVID-19 spread across the US
AI Frame Interpolation Boosts LEGO Stop Motion to 60 FPS
It takes a whole lot of work to create stop-motion animation. After all, the characters and scenery in each and every frame need to be precisely positioned before snapping each still image. At the traditional film rate of 24 frames per second, you need to shoot 1440 individual photos for one minute of footage. That’s why stop-motion animators sometimes choose to go with even lower frame rates, like 15 fps.
Digital animation tools usually include something called “tweening,” which can interpolate illustration positions to come up with an in-between frame. This can smooth out motion in two-dimensional images, but this approach doesn’t work well with photography. Now, thanks to neural network technology, it’s possible to cleanly interpolate frames between photographic images with impressive levels of accuracy.
Recently, LEGO stop-motion animator LEGOEddy decided to run one of his 15 frame-per-second animations through a tool called DAIN, which was able to upsample his original video to a super-smooth 60 frames-per-second. The software not only interpolates frames but is able to properly handle things like depth of field and occlusion (objects being hidden behind others.)
The resulting footage is impressive, and surprisingly devoid of unpleasant artifacts. Here’s the full 60 fps version of LEGOEddy’s “Apollo 11: A Lego Story”
And here’s the original 15 fps version for comparison:
You can find out more about DAIN on Two Minute Papers, or download the application for free (or for a small contribution) from the project’s Patreon page, or the source code on GitHub.
[via Born in Space]