AI versus Marshy - Midjourney solves the continuity problem
I’m Marshy, a content creator who’s been experimenting with AI tools to enhance my work. I’ve been sharing my experiences and insights with you through this newsletter, and today I want to talk about a specific challenge I’ve been facing with image generation using Midjourney. If you’re new to this newsletter, don’t worry - you can still follow along and learn about the latest developments in AI. Hello and welcome to AI versus Marshy #39! Well isn’t this the little engine that could! I’m currently sitting in an airport waiting for my plane home, after doing a video shoot for a bank, working from the startup space Fishburner’s, eating too much gelato, and attending a HR tech conference. But enough about me - let’s talk AI! This week looks at: Midjourney solves the continuity problem If you were starting out as a content creator today - how would you use AI? Another ADHD project update Lots to cover so let’s make like an envelope and letter it out*. -Marshy Repeat your character Via VentureBeat . I’m sure most of you have played with text-to-image at some point by now. If not, stop reading and sign up for Adobe Express. It will let you generate a text to image for free. I just did this: I want you to draw an excited business man in a grey suit with a basketball singlet over the top and backwards cap, riding a comically small rocket with red and blue and yellow flames sparking out the back while it rises at a 45 degree angle. The style should be pen and water colour and with a limited 16-colour palette. One of the options presented this amazing character who I am quietly impressed with: Business baller is amazing. But here’s the rub. Say I wanted my business baller guy to hit the basketball court, or go for a ride on his unicorn, or start smashing some pasta and spaghetti sauce with glee - I can’t. The short answer is because it’s using a “diffusion model”. These run predictively from text - and it’s mind-bonkingly hard to get the continuity that we take for granted. Midjourney is one of the most powerful image generators out there right now and is playing with solving this by adding a toggle called “-cref” - which in layperson’s terms means: Originally appeared in newsletter : AI versus Marshy #39: continuing characters, content preso, and another ADHD update
Want more of this?
Weekly-ish thoughts on AI, growth, and being human in tech. Sometimes useful, sometimes not.
Subscribe to AI versus Marshy →