Gan Dreems VJ Pack 1 - Nature
Welcome to my Gan Dreems VJ Packs!
GAN stands for generative adversarial networks and represents a class of machine learning frameworks. Once trained on a set of images, the network can then generate images that have similar statistics to the training set. This results in a huge set of frames with a combination of properties similar to the training set. By interpolating between specific keyframe values, we get abstract videos with an incredibly smooth and fluid flow between them. That's my oversimplified explanation, but I encourage you to go learn more about this fascinating code.
All loops have been rendered using the BigGAN model
All loops have been upscaled using different AI upscaling methods depending on what method looked best for each loop.
All loops are 30 fps, 1080p X 1080p so mixing with additional effects is encouraged to get a nice FHD output (see video for examples). But feel free to use/mix these however you want.
Most of these loops are seamless and have the same start/end frame.
Most of these loops are 20s long.
Promo video is broken into 3 parts:
1. All loops displayed on screen together at normal speed but tiny resolution
2. All loops scrolling across the screen at increased speed but full resolution
3. Actual loops mixed in Resolume with additional effects to show some VJ examples
Music from the video by my good friend Maxfield. You can listen more here:
https://soundcloud.com/jake-maxfield
If you have any questions you can email me at: