Creativity AI #23: Weekly dose of AI art
AI news, Midjourney update, Prompt Play, Landscape lighting
Creativity AI is a new publication that brings together interesting articles and recent developments in the AI world related to creativity and productivity.
AI News
Midjourney Office Hours updates
What to expect this month
New Style References releasing soon (within a week)
Updated Style References with improvements
Initial video models (image-to-video prioritized)
Version 7.1 update (possibly)
Next Month
Style Explorer features launch
Server Infrastructure
New servers being set up for video model support
Additional expansions planned for following months
Version 7.1 Details
Improved body anatomy rendering
Better hand coherency
Video Models
No Relax mode initially (to prevent server overload)
Relax mode possible for higher-tier users later
Basic functions and short clips first
Longer videos coming after launch
Annual subscribers only initially (funding for GPUs)
40% of users have annual subscriptions
Two versions planned: budget and professional-grade
Style References Updates
Enhanced style reference accuracy
Random style support added
Works with Moodboards
Omni Reference
Improvements below expectations
Further enhancements delayed until Version 8
Runway released "Layout Sketch," which allows users to guide image composition by drawing on a blank canvas or directly on top of an image.
Let's test it out. Here is my prompt.
a woman scientist wearing a white lab coat and blue frame glasses is holding a roseThis is my beautiful drawing using Microsoft Paint:
The result:
I also tried using the same image (without the sketch).
The result is not what I expected because (1) the woman holds the rose differently, (2) the flower stem is missing, and (3) the woman holds the flower with both hands.
Perplexity has released Perplexity Labs, which can perform a variety of tasks for you even with a simple prompt. It can perform a number of automated analyses, including infographics and tables. You can see what others are doing with the Labs here. One notable creative use case is creating storyboard images for a story (see an example from the community).
Unlike ChatGPT-4o, Perplexity Labs create multiple images for a storyboard with just one prompt.
This is my prompt for Perplexity Labs:
Write a short story about a mother cat surviving in a dystopian science fiction world. The cat must find food or her kittens will die from starvation. She is in a race against time. Create 9 storyboard and a full screenplay.It generated the stories and storyboard images.
Here’s an example:
Microsoft offers a free course (21 lessons) about generative AI for beginners. You can find the resource here.
Claude also provides a free training course called AI Fluency. It focuses on human-AI collaboration rather than understanding AI as a technology. You can find the resource here.
If you're wondering if OpenAI ChatGPT offers educational content, you're in luck! Yes, they've created their own learning hub here with a wealth of valuable information straight from the source.
Update from Geeky Curiosity
Midjourney Style Resources #6: Dot art, pixel art, dot painting, dot illustration, pointillism, stippling has been updated to V7.
Midjourney Style Resources #6: Dot art, pixel art, dot painting, dot illustration, pointillism, stippling
This article is part of the resource series "Geeky Curiosity's Boutique Collection of Midjourney Styles".
Prompt Play
Today's Prompt Play is an interesting experiment, but the outcome is rather surprising.
Remember last week's Creativity AI #22, where Ideogram gave users a prompt template to create a logo? Okay, maybe last week was a little too long to remember (joking), so I'm pasting the prompt again here:
The text "{Brand Name}" is written in a {clean / bold / elegant / playful} {serif / sans-serif / cursive / handwritten / geometric / pixelated} font, [with {optional color} outline,] [rendered in [lowercase / uppercase / small caps}, and] positioned {below / beside / integrated with} the logo[, with {optional tagline} in smaller text]. Like many things in the AI space, I strongly advise you NOT TO TRUST what the developers say until you try it for yourself. Sometimes, the claims are for super polished results (perhaps one success case per 100 tries) or hidden flaws that they don't want you to see.
Anyway, I tested it out following my recommendation to see if the AI models can accurately generate fonts in a variety of styles.
I simplified the above template to experiment with six different font styles: serif, sans-serif, cursive, handwritten, geometric, and pixelated.
Serif fonts feature small decorative lines or "tails" (known as serifs) at the ends of strokes within each character. "Sans" is a French word that means "without," so sans-serif is similar to Midjourney --no parameter to make sure the font does not have extra decorations.
ChatGPT-4o
These are the prompts for ChatGPT-4o:
Create an image for this: The text "Typography" is written in a serif black font on a white background. Create an image for this: The text "Typography" is written in a sans-serif black font on a white background. Create an image for this: The text "Typography" is written in a cursive black font on a white background. Create an image for this: The text "Typography" is written in a handwritten black font on a white background. Create an image for this: The text "Typography" is written in a geometric black font on a white background. Create an image for this: The text "Typography" is written in a pixelated black font on a white background. The results:
ChatGPT-4o passed with flying colors. Except in one case, it created a handwritten font as a serif font.
Ideogram V3
These are the prompts for Ideogram V3:
The text "Typography" is written in a serif black font on a white background. The text "Typography" is written in a sans-serif black font on a white background. The text "Typography" is written in a cursive black font on a white background. The text "Typography" is written in a handwritten black font on a white background. The text "Typography" is written in a geometric black font on a white background. The text "Typography" is written in a pixelated black font on a white background. The outcome is shockingly bad. Except for one, none of the texts are correctly rendered. The rest were all typos. The font styles are fantastic, but they are ineffective if the text is misspelled. I believe the word "typography" is too long for a 1:1 aspect ratio image. So I tried a 3:1 ratio (long horizontal canvas), but Ideogram still failed miserably.
Only one image is usable out of repeated 2-3 times per generation (4 images per font type).
Conclusion
ChatGPT-4o is a clear winner in terms of text accuracy and font style.
Inspiration
I was inspired by beautiful landscape photographs on the internet. So I’m trying out some new keywords for landscape photography and let’s see what Midjourney V7 could do.
This is my overall Midjourney permutation prompt, testing out 10 light keywords related to landscape.
landscape photograph of a small hill near a river, {light spill, catchlight, continuous light, low key light, high key light, penumbra, form shadows, twilight, golden hour, reflected light} --ar 16:9 --v 7Light spill
Light spill describes how light spreads over an area beyond the intended target.
landscape photograph of a small hill near a river, light spill --ar 16:9 --profile r3snb8i --v 7
Catchlight
Catchlights in landscape photography can appear in water surfaces, wet rocks, or ice, adding sparkle and visual interest to compositions.
landscape photograph of a small hill near a river, catchlight --ar 16:9 --profile r3snb8i --v 7
Continuous light
landscape photograph of a small hill near a river, continuous light --ar 16:9 --profile r3snb8i --v 7
Low key light
landscape photograph of a small hill near a river, low key light --ar 16:9 --profile r3snb8i --v 7
High key light
landscape photograph of a small hill near a river, high key light --ar 16:9 --profile r3snb8i --v 7
Penumbra
Penumbra is the transitional area between full shadow and highlight.
Form shadows
Twilight
landscape photograph of a small hill near a river, twilight --ar 16:9 --profile r3snb8i --v 7
Golden hour
landscape photograph of a small hill near a river, golden hour --ar 16:9 --profile r3snb8i --v 7
Reflective light
landscape photograph of a small hill near a river, reflected light --ar 16:9 --profile r3snb8i --v 7
Recent articles
Related articles
Free Geeky Animals' Sref code: 3595044843
Cover prompt: Elegant art nouveau illustration featuring a graceful lady with a cat, surrounded by swirling floral vines and soft botanical motifs --ar 16:9 --sref 3595044843 --profile r3snb8i --sv 4 --v 7
I hope you like this article!
Thank you for reading and happy creating!