Creativity AI #16: A digestible weekly dose of AI updates
Midjourney Office Hours (2025-04-16), AI news, prompt play (ChatGPT-4o vs Ideogram 3.0), and inspiration (free Happy Easter Midjourney V7 prompts)
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 (2025-04-16)
Midjourney V7 will leave the Alpha stage in 30 days.
These are the upcoming releases:
(1) Today or tomorrow
Editor updates for improved image selection and layers. The new editor will not need a yearly Midjourney membership
A new version of the --weird parameter
Fast mode, which is both faster and cheaper. (Relax and Turbo modes will also become faster.)
Aesthetics and Personalization will be updated, including a slight improvement to the base/default model's anatomy. There may be a rating party to improve the default aesthetics.
(2) Probably next week
A new quality parameter (--q 4) that is slower but has better aesthetics.
(3) Probably 2-3 weeks
Omni-referencing (improved subject and object references)
A new "Super Style References (Sref)" that is better than incremental improvements. The new Sref will include more features. It will also improve the moodboard.
Better Upscalers
(4) Unknown timeline
Ultra-fast Draft mode is 5 times faster (nearly instant speed).
David believes that a more aggressive approach to aesthetics and fine-tuning must take into account trade-offs, such as prompt following and image quality.
Personalization benefits 80-90% of people. However, some people (including Midjourney employees) dislike it. But Personalization will get better by tomorrow. Or, try rating more to get a better Personalization profile.
The Explore page may include a Personalization page (for exploring community-shared P codes).
The video model will complete its training by the end of the month. Then, it will go to the next phase. Midjourney video will be very different from others on the market. It will be somewhere in the middle of quality and price.
Faster models will allow developers to explore new features in the coming months.
There may be V7.1 or V7.2 for improved image quality, but this is still being discussed. David talked about the Midjourney V8 and whether to bring all of the features to V7 or go straight to V8.
According to David, there is currently no good way to add the "multi-prompts" feature to V7. ( "It's not a quick thing" ). It is the same reason why the "--no" parameter does not work well in V7.
Moodboard is style-constrained and focuses on style. Personalization focuses on broader subjects and is more situational.
Midjourney V7 is better in the following:
Create spaceship and sci-fi images
Blending things
Things are not in Midjourney V7:
Not focusing on transparency
There is no newer version of Describe
Why is V7 bad at text? It is because the text was removed from the training dataset.
Retexture on some parts of the model is possible, but the developers do not have the time to develop it right now.
David says "Midjourney" because we are "mid-journey" with a rich and unimaginable future ahead of us as well as a rich history behind us.
News
Ideogram allows its users to limit the number of generated images per grid, from 1-4. It allows you to save some (fast) credits and get to what you need faster. I think it should be quite useful, especially if you have already optimized the prompt and are simply changing the subject/object to generate more variations.
Perplexity now offers Gemini 2.0 Flash Experimental, in addition to the existing FLUX.1, and DALL-E 3 image generation models. You can set the default image model at Preferences to start creating images using Gemini. For example, you can type this prompt to create an image: "Generate a photo of a happy bunny celebrating Easter."
What you should know about the images created with Gemini 2.0 Flash Experimental in Perplexity:
Perplexity generates text. If you want to create images (seriously), use another AI tool.
1 image per generation.
The image made by Perplexity has no watermark, unlike the image created in Google AI Studio.
The image quality is significantly lower than Midjourney.

Google published a (free) whitepaper outlining prompt engineering techniques and best practices for becoming a better prompt engineer. Some of the topics covered in the 68 pages document include zero and few shot prompting, system prompting, role prompting, contextual prompting, step-back prompting, chain of thought, self consistency, tree of thoughts, and ReAct. Download a free copy here.
Adobe will roll out an AI agent to improve the creative workflow in Photoshop, allowing the bot to suggest areas for improvement in the image with a single click. Adobe intends to include over 1,000 one-click actions in Photoshop to speed up editing and assist users in learning and using the software.
According to its website videos, the new image editing process is "more clickable" and interactive. Clickable helps because it enables you to type less, like when editing images in ChatGPT-4o. But it’s not yet available when I checked my Photoshop.
ElevenLabs makes cloning your voice easier with its new Professional Voice Clone (PVC). By default, your PVC is private and only you can access it. But you can monetize it if you want, and join thousands of other voice actors who earn royalties every time the user generates audio using your voice clone.
Canva announced its new Visual Suite 2.0, where "Productivity meets Creativity." It claims to have all of the tools you need to manage your workflow in one place. In their keynote video, they demonstrate how to use Canva Sheet to create visually appealing charts and graphs, use AI to bulk-create social media content and create an interactive app with Canva Code.
However, this one has a red flag. All of these new features are "coming soon" - so I couldn't test them to know if they'll be as good as they claim or if they're just hype. We'll see in the near future.
OpenAI released three new models for its API: GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano. The newer models outperform GPT-4o and GPT-4o because they have larger context windows (up to 1 million tokens of context), which allows for better long-context comprehension. They have strange names because ChatGPT-4.5 was released some time ago, and now we're back to ChatGPT-4.1, which is supposed to be better? Anyway, these models are not yet available in the ChatGPT website and user interface that most users are familiar with.
Prompt Play
Ideogram is well-known for its mastery of typography. Indeed, it's my go-to AI tool for creating text and typography artwork. But what if I compare it to ChatGPT-4o that is also known to be the expert in text generation?
Which tool is better at creating a set of fonts?
This is my prompt for creating a font set.
a complete full character set: "A B C D E F G H I J K L M N O P Q R S T U V W X Y Z 0 1 2 3 4 5 6 7 8 9 " in the style of science fiction on plain white background.
For ChatGPT-4o, I used the "Create Image" button and pasted the prompt over so I didn't have to reformat it in the conversational style.
For Ideogram 3.0, I kept the Magic Prompt enabled (the results are generally better because the bot will help to improve the prompt).
The prompt was used four times in ChatGPT and Ideogram to ensure that the results were consistent and not purely chance.
The results:

Summary of the findings:
ChatGPT accurately generated the alphanumeric characters. But it also repeated the numbers at the end.
Ideogram 3.0 generates characters that look better than ChatGPT but are less accurate. The generated alphanumeric characters are incomplete.
ChatGPT-generated characters can be subsequently enhanced with Midjourney Retexture to create font variations.
I also changed the prompt to include special characters and detailed visual descriptions, but neither AI tool could generate the entire character set (result not shown).
In conclusion, ChatGPT-4o outperforms Ideogram 3 for quickly generating a font set in a single image - which is quite surprising and unexpected.
Ideogram's font looks better but less accurate. The fonts (both ChatGPT and Ideogram) could be further enhanced using Midjourney Retexture.
Inspiration
It's almost Easter. Here are some of my prompts that will help you create some lovely Midjourney images or e-cards for your family and friends!
(New to Midjourney V7? Read this article to set up your V7 right from the start. It includes the tips to prevent frustration and maximize your creative potential.)
Wishing you a very happy Easter!









Recent articles
Related articles
Free Geeky Animals' Sref code: 2076092050

Cover prompt: crocodile-headed divine figure in golden-lit chamber, magical pouch bursting with gold bars, reflective surfaces, divine glow --ar 16:9 --sref 2076092050 --profile vumoqva --sv 4 --v 7
I hope you like this article!
Thank you for reading and happy creating!