Creativity AI #17: Weekly update that doesn't overwhelmed
Midjourney Office Hours (2025-04-23), Midjourney Magazine Year One, AI News, Prompt Blending, Eudemonic Well-being
Creativity AI is a new publication that brings together interesting articles and recent developments in the AI world related to creativity and productivity.
Midjourney Office Hours (2025-04-23)
The next release is most likely two weeks away. It's not sure if there will be any releases this week.
The developers are currently working on fast mode, --q 4 (better quality), omni-reference, better image quality, better personalization to improve aesthetic, and an improved editor (particularly for non-Midjourney images).
The goal now is to improve the images using the default base model.
It's said that the image quality for V7 could be 10 times better. However, the developer will need a tool to accomplish this. Some new image settings may improve image quality (conservative setting), but non-conservative settings may cause prompt adherence to break. The developers believe that the new experimental settings should be released, even if they do not yet have appropriate names.
David mentioned Midjourney V8 multiple times during the Office Hours. It seems that things are moving quickly toward V8, even though certain features (such as inpainting and retexture) from V6.1 might not be made available in V7. The 3D stuff will be in V8. David wants to run the V8 from the new servers.
There is a "crispy burning image artifact" in V7 that the developers are currently working to fix.
The Midjourney video shows the "first sign of life," and the developers were pleasantly surprised by the quality of the videos. However, they still need to be trained and fine-tuned for the coming weeks. The Midjourney video is affordable and consistently good, with the Midjourney aesthetics.
The Niji team is working on their video model.
Update from Geeky Curiosity
Midjourney sent me their book "Midjourney Magazine, Year One." The book is a collection of twelve issues of Midjourney Magazine.
The best part? It's free because Midjourney says, "You are one of the most dedicated users in the Midjourney community. As a thank you, we'd like to send you a copy of our book, Midjourney - Year One." :-)
The images generated in legacy systems (V5 and V6) still look gorgeous, though they cannot compete in terms of image detail with V7. But AI-generated art, like all art, is timeless.
The book is keyword-rich and a great resource for experimenting and learning new keywords in V7 and beyond.
News
Perplexity has added the latest GPT-4.1 model, which supports up to 1 million tokens. Perplexity also introduced its iOS Voice Assistant.
Vercel's survey (Q1 2025)of app builders reveals that 87% of developers use OpenAI, 68% use Anthropic, and 63% use Google, with Microsoft scoring the lowest at around 10%.
OpenAI is still the most popular LLMÂ model among developers. The primary application of AI is for customer-facing applications (43%), with a focus on product AI, knowledge base, and Q&A.
The top technical challenges in developing AI features are accuracy and hallucinations (approximately 60%), latency/performance, and cost management.
Runway is offering 20% off the first month of its Unlimited Plans to all new subscribers until April 23rd, 2025. It's a good deal if you want to test out their Gen-4 video model.
Google is now offering free video generation using its Veo 2 model in Google AI Studio. The feature set is somewhat limited (fixed aspect ratio, 8-second video duration, 24 frames per second, 720p resolution), but it is free, so I'm not gonna complain.
This is video clip I generated using Veo 2 based on a Midjourney V7 image.
Prompt Play
Prompt blending is one of those techniques that I enjoy playing with, but I'm not sure what to do with it or how to turn it into something practical and useful.
It's quite simple. First, create at least two unrelated prompts and then combine them with or without the "+" sign. And you let Midjourney or another image generator work their magic to blend those prompts. The results can be creatively interesting or result in a split image because the bot has yet to figure out how to mix them!
Here's a quick prompt blending experiment that compares ChatGPT-4o and Midjourney's blending capabilities.
Test 1: The Dog and Snowman Family
ChatGPT prompts: (putting two different prompts together)
A puppy wearing a scarf jumping into a pile of autumn leaves + A snowman family posing for a winter holiday photo. Aspect ratio 16:9
The bot failed because it couldn't figure out how to blend them. So, we ended up with two images stitched into one. However, this could be a hack if you want to create two different images in one go because ChatGPT's image generation can be extremely slow.
Now, we try again with Midjourney V7:
A puppy wearing a scarf jumping into a pile of autumn leaves + A snowman family posing for a winter holiday photo. --ar 16:9 --profile vumoqva --v 7
Wow! Midjourney is the clear winner in this one. Nice blend result with plenty of action.
However, it did not always work. Simply re-roll and try again.
Test 2: Baby Dragon and Cloud Sheep
Let’s do another test. Here’s the ChatGPT prompt:
A baby dragon trying to toast marshmallows with its tiny flame + A cloud sheep dozing off while floating in the starry sky. Aspect ratio 16:9
ChatGPT simply placed these two prompts together without much fanfare.
Compare it to Midjourney.
A baby dragon trying to toast marshmallows with its tiny flame + A cloud sheep dozing off while floating in the starry sky. --ar 16:9 --profile vumoqva --v 7
Ahh, much nicer results. The piece appears to be complete and not assembled arbitrarily.
So Midjourney wins again!
So, next time you're bored or out of ideas, try prompt blending! Put those unrelated prompts into Midjourney, as if it were a giant blender, and mix thoroughly. You may be pleasantly surprised by what you will get!
Does looking at art make you happy? "Yes," said the researchers.
We don't need scientific evidence to prove that art makes us happy and helps us navigate and cope with our demanding and often chaotic modern lifestyle.
Yet, there are many papers published on this topic. The results of 38 studies (N = 6,805 participants) show that viewing visual art (e.g., paintings, art posters, sculptures, mixed media) can improve "eudemonic well-being," which refers to a sense of meaning, purpose, and fulfillment in life.
There are five mechanisms for how art viewing improves well-being:
Affective processes involve evoking positive emotions, pleasure, and emotional regulation.
Cognitive processes involve enhancing attention, memory, exploration, and learning.
Social processes include improving social connections and reducing loneliness/isolation.
Self-transformation involves promoting reflection, empowerment, identity affirmation, and a sense of purpose.
Resilience promotes psychological restoration, coping skills, and healthy behaviors.
Art viewing is a low-threshold, easily accessible activity that can be done in a variety of settings, including museums, hospitals, workplaces, online, and at home. It requires no special skills or equipment and can last as little as a few minutes, making it a simple and adaptable tool for promoting mental health and well-being.
With Midjourney, Ideogram and other AI art generators, you too can be an artist and create artwork that resonates with you. Ignore the AI-art haters, create something that makes you happy, and take care of your well-being.
Inspiration
Inspired by the new term "eudemonic well-being," I used the MJ Simple Prompt Generator to interpret the academic terminology and suggest some different aesthetic directions.
The results are as follows:
My conclusion is that the keyword "eudemonic" is a little too "demonic," so the bot associates it with evil and demonic creatures. The phrase "eudemonic well-being" helps the bot understand and generates more images related to well-being.
By entering the phrase "eudemonic well-being" into the MJ Simple Prompt Generator, the visual interpretation of the phrase deepens and diverges as ChatGPT translates its meaning and generates prompts that lead to various aesthetic directions.
Recent articles
Related articles
Free Geeky Animals' Sref code: 3586743350
Cover prompt: female cyborg enhancing herself with a golden biomechanical arm in a neon-lit laboratory, glowing tech, high-tech sci-fi details --ar 16:9 --sref 3586743350 --profile vumoqva --sv 4 --v 7
I hope you like this article!
Thank you for reading and happy creating!