Year: 2022

Now. How would Helmut Newton photograph the Medusa?

Like this. At least according to #stablediffusion2.

(again, no inpainting or photoshop of any sort) https://t.co/Gxx6TNDFXq

RT @fosta: Westfield Mall, San Francisco, noon, the Saturday before Christmas, 2022. https://t.co/t3VkVkGrCA

Even if it seems impossible, the pace of #GenerativeAI is accelerating. *Every day* now, there are at least 2 new techniques or tools developed by the community that I must try to stay on top of things.

I’ve never seen anything like this in any field in my 23 years in IT.

Maybe this comment by @EMostaque was not registered by the #AI community:

“A MacBook M2, as of today, can generate an image in 18s, of any type. In a couple of weeks, it’ll be less than a second”.

Subsecond image generation before the end of 2022.
What will happen in 2023? https://t.co/XBKDdHKL4T

@paulg Using ChatGPT to come up with new Pokemon ideas, SD to generate visuals based on the descriptions and a bit of Photoshop to bring it all together
https://t.co/ZLdiwNCUuj https://t.co/s5mLygAqwB

@jfischoff Dream Camera. You can join the beta, follow the conversation, or talk with the developer here: https://t.co/WTAd1oUUzH

@rainisto Understood. I was more after your personal perception if synthetic media is valued more, equal, or less than non-synthetic media. (The market might have a different opinion)

@rainisto That’s interesting. I’m not familiar with the whole process so I’m going to ask a potentially stupid question: can you decide the price? If so, did you price these same, more, or less than non-synthetic pictures?

@TomLikesRobots @nptacek @EMostaque What we really need to figure out is how to generate the “back” portion of the object in a way that is both consistent with the front and merges with it gracefully.

RT @TomLikesRobots: Simple Stable Diffusion to Blender:
1. Use Automatic1111’s depth map extension (https://t.co/3i4QhbvAV8) to create imag…