June 2023

I had an hour play around on the new LaLamu Studio demo, there’s a time limit and crunchy quality for now, but it works as an example of the speed and easy access everyone will have to these tools before long. It’s been possible to add a moving mouth to images for a while but it’s new with video. There will be amazing advancements for animators and video editing, but also lots of worrying possibilities for proving anything is real.

You select an audio clip (even uploaded audio) and a video of a face and in 30-60 seconds you have a clip of that face speaking the words. You can upload video and audio (with limits) to test it.


January 2023

Snoop Dogg & Matt Berry: Barbie Girl (cover) – (apologies for my crimes against music)

A quick experiment I made a couple of weeks ago, using the AI voice generator uberduck and some audio/video bits chopped together, I had to mess up the video to get past copyright. Developments in publicly accessible AI models are happening so quick at the moment, and lots of voice/people cloning is happening too and isn’t difficult. I’ve been following it a bit and will be posting some of it here. We are in an era of computer generated content, with chatgpt written songs and books, prompt-to-image artists and a whole new world of copyright claims.

August 2022

First proper test of Stable Diffusion as a source of images to make video from.

This took ages, a learning process. I’d found the horse/dinosaur/old photo mix giving good textures and quality when playing around, so explored it further using long complex text prompts with small changes each time. End result is a bit dark, a bit twin peaks influenced, especially with the audio. The sound was mostly from a squeaky metal gate in Arminghall among other things. 

August 2022

Johnny Vegas makes giant teapots, calm -> wild

I’ve been testing out the beta for #stablediffusion , one of the new #aiart models emerging. I think I’ve spoke about it to everyone I’ve seen. It blows my mind every time. These are images generated just from word prompts, similar to the #dalle2mini thing people were using recently.

A few days ago I had listened to an old RHLSTP podcast where Michael Pennington was talking about being Johnny Vegas, his enjoyment of making pottery, and how he used it as part of Jonny’s character. I thought I’d develop it further for him and let him make some really big teapots. I tried lots of prompts and variables to get these. Will post some other more creative experiments too.