The Checkpoint #6 – Scene-stealing blobs and glitchy symphonies

News & articles

Latent Princess Generator – Not just for princesses, this is a new general-purpose Clip-guided latent diffusion approach from @dango233 @multimodalart which produces excellent results.

Centipede Diffusion update – I've featured this notebook a lot, but it's just so good. Latest release sees a new UI for precisely selecting where to inpaint. Definitely worth checking out.

Thin-Plate Spline Motion Model for Image Animation – Cryptic name, but can transfer motion from a video to a still image startlingly well – Colab, Github, Paper

SymphonyNet – Generate symphonic soundtracks to compliment your AI movies. Pretty cool, though the results do get more mad as they progress.

knylx @knylx_art

To my fellow AI artists: A lot of critics are going to tell you that you are not a 'real artist'. When was the last time in your life that you were a part of the avant-garde, we are at the beginning of a revolution and you are part of it. Be proud. (1/3) photo: @GANPRSN https://t.co/O5C2c08ey6

Featured notebook – BlobGAN

Instead of generating an image as a whole, BlobGAN constructs scenes from depth-arranged 'blobs' of elements. The upshot is that by moving and resizing a blob, the network moves and resizes a wardrobe in your bedroom scene, for example.

The interface for editing blobs is tricky and the ability to upload custom scenes hasn't been released yet, but it works well as a first release. This kind of feature built into something like Photoshop would be huge.

Get it

• Website

• Github

• Colab

Thanks for reading!

If you have anything you’d like to be featured or want to get in touch, give me a shout on Twitter or send an email. Please also consider supporting me on Patreon so that I can spend more time creating content like this.