Creativity AI #35: Weekly update for AI art
News about AI art, Changes in Geeky Curiosity, Prompt Play, Inspiration
Creativity AI is a free weekly publication that brings together interesting articles and recent developments in AI art and creativity.
You can also read this newsletter on the website.
News about AI art
Midjourney's latest office hours talked about new video capabilities, with the HD video mode (720p resolution) finally and officially rolling out after much anticipation. It was already rolled out to some users weeks before.
Unfortunately, the HD video feature is exclusively available to Pro and Mega tier users due to the hefty costs, and there's no relaxed mode option. What's particularly frustrating is that you can't test your video at standard resolution first and then upgrade to HD if you like it. You're essentially gambling with your credits each time since you won't know what you'll get until you commit to the expensive HD generation. But this approach is not unique to Midjourney and quite common among other AI video generators in the market.
Upcoming releases (end of August):
A new style Explorer tool
Possible partial v8 integration into version 7.1 (though success rate is only 50/50)
Current and potential features:
HD video mode available now (manual activation required in settings)
Turbo mode for super-fast video generation (uncertain release, probably not going to happen)
2.5-second video mode as an alternative to the standard 5-second clips (uncertain release)
Plans to introduce a dedicated tab for Moodboard to improve discoverability (not sure yet)
The V8 launch is still several months away as the developer prioritizes quality over speed.
Example of the (expensive) HD video. Tried to cheat using /Relax mode but no. Can’t.
For the first time, OpenAI just dropped free AI models you can run on your computer. No more subscriptions or API fees. The gpt-oss-120b matches o4-mini's intelligence and runs entirely on your own hardware. Sure, it's text-only (no images, videos, or audio), but this is still massive.
What you can do:
Run AI completely offline on your laptop
Execute code and web searches
Use chain-of-thought reasoning
Adjust processing effort for tasks
Build AI into mobile apps (20b model)
Want to actually use these free AI models? Just grab Ollama (free too) that makes the installation easy-peasy. Download it, click install, and you're running OpenAI's models locally in minutes.
Anthropic introduces the upgraded Claude Opus 4.1 model, and I'm excited about this one. Last week, I was stuck on a nasty software bug and ChatGPT o3 couldn't help me crack it after multiple attempts. Switched to Claude Opus 4.1 and boom, solved in two tries!
Key improvements over Opus 4:
200K context window for complex tasks
Real-world coding that actually works
Advanced reasoning and agentic search
Natural, prose-focused creative writing
What impressed me most? It doesn't just write code, it understands the problem. For complex “vibe coding”, this feels like having a senior developer on call.
Google just integrated LearnLM into Gemini with "Guided Learning" – turning it into an AI tutor that actually teaches, not just answers. While ChatGPT launches Study Mode and Claude for Education targets classrooms, Google's making its move in the AI education race. What comes next? Probably more tools that help you genuinely understand concepts instead of just memorizing facts.
What you can do with it:
Break down complex problems step-by-step and get explanations adapted to your level
Generate flashcards, study guides, and interactive quizzes from your materials
Access automatic visual aids like diagrams and relevant YouTube videos for better understanding
Ideogram is really enjoying itself with the Character feature. It recently introduced the Templates feature to make character consistency more accessible to all. Now's the time to do a Face Swap using an uploaded image or meme! It's really easy to use and a blast to see your character's face in all sorts of situations. Manually writing a prompt is not needed when it’s used with the templates.
ElevenLabs just dropped their Eleven Music model that lets you generate songs and music from simple text prompts, and you can use them commercially! Now, I won't sugar-coat it: this isn't quite at Suno or Udio's level yet.
It feels more like a bonus feature tacked onto your existing voice and sound effects subscription. Still, pretty cool to see them branching out beyond just voices. Worth experimenting with if you've already got their subscription running.
Scientists just captured something incredible: wild eastern quolls (Dasyurus viverrinus) glowing electric blue under UV light! Photographer Ben Alldridge documented this biofluorescence phenomenon for the first time in Tasmania's wilderness.
Yeah, I know, I know. It looks like an image generated with Midjourney, but it’s not! That is an actual animal!
While these marsupials normally look brown with white spots, UV light makes them light up like disco balls. This discovery is helping researchers understand light pollution's impact on wildlife.
Read more: Nature's best science images
Update from Geeky Curiosity
I'm bringing text and music generation AI news back to the weekly updates—figured you'd want the full picture of what's happening across creative AI tools, not just the visual stuff.
Leave a comment and let me know if you’re happy with this change.
Refreshed Midjourney Style Resources
The Midjourney Style Resources #13 has been updated to V7 aesthetics.
Prompt Play
I tried five AI image generators to create a simple book cover. The result is not exactly what I expected.
ChatGPT is the reliable choice
ChatGPT nailed it. Clean book cover, correct 5:8 aspect ratio (perfect for standard book dimensions), and it actually understood what I wanted. The cat looked adorable, the text was readable, and everything felt book-ready. No surprises here.
a book cover titled "Cute Kitty Cat" featuring a cat holding a coffee mug, dreamy background with stars and hearts, soft colors, children's storybook cover. Aspect ratio 5:8
Midjourney V7 did it well this time
Here's what caught me off guard: Midjourney generated the text perfectly. Could it be the developers secretly improving the text behind the scenes? The book cover looked great, maintaining the children's storybook vibe and following the prompt precisely. Another solid 5:8 aspect ratio without any fuss.
a book cover titled "Cute Kitty Cat" featuring a cat holding a coffee mug, dreamy background with stars and hearts, soft colors, children's storybook cover --ar 5:8 --profile uw4q9o7 --v 7 
Ideogram 3 didn't deliver this time
I had high hopes since Ideogram usually handles text well. And yes, the text looked decent. But somehow, it decided I wanted the entire book instead of one book cover. Plus, there's a background that I didn't ask for. The bot doesn't understand the user intent.
a book cover titled "Cute Kitty Cat" featuring a cat holding a coffee mug, dreamy background with stars and hearts, soft colors, children's storybook cover. Aspect ratio 10:16
Runway Gen 4 is the video generator trying its best
Among the generated images, several showed a cat holding something map-like. But I chose this one that looks cutest. The quality is there, just didn't scream "book cover." The main issue is that the aspect ratios are basically non-negotiable. You get a set menu and the 3:4 is the closest. Not ideal when you need specific book dimensions.
a book cover titled "Cute Kitty Cat" featuring a cat holding a coffee mug, dreamy background with stars and hearts, soft colors, children's storybook cover. Aspect ratio 3:4
Kling KOLORS 2.1
Kling KOLORS 2.1 is Kling’s latest image generation model. It generated an adorable children's book style with a sweet cat holding a mug. The artistic style may appeal to some publishers. However, like the other video generator Runway, it's still limited by fixed aspect ratios. No 5:8 aspect ratio option.
a book cover titled "Cute Kitty Cat" featuring a cat holding a coffee mug, dreamy background with stars and hearts, soft colors, children's storybook cover. Aspect ratio 2:3
My take
For actual book covers, stick with ChatGPT or Midjourney. Both understood the assignment, delivered the right dimensions, and created usable results.
Video generators like Runway and Kling? Save them for their intended purpose. The aspect ratio limitations alone make them impractical for book cover work.
Ideogram 3 needs to chill with the extra interpretations. A book cover is just a book cover.
Inspiration
Evolution isn't the one-way street we thought it was. The tomatoes in the Galapagos Islands prove this beautifully.
We've all learned that evolution pushes species forward, right? Better adaptations, more specialization, survival of the fittest and all that. But these tomatoes? They're basically hitting the rewind button. They're producing the exact same toxins their ancestors made millions of years ago.
Talk about a plot twist.
Scientists call it "reverse evolution," and honestly, it makes perfect sense when you examine the environmental pressures. These plants faced a specific challenge on the islands - the soil is less developed and the landscape more barren than what modern tomatoes typically thrive in. So what did they do? They reached back into their genetic toolkit and pulled out an old trick that worked before.
This whole concept sparked something in me creatively. What if we could visualize this backward journey? What would it look like to capture evolution running in reverse? That question led me to create these images - my attempt to explore what happens when nature decides the old ways were actually the best ways.
a digital human face morphing backward into apes and early hominids, each stage regressing in complexity, set against a decaying circuit-board background --ar 16:9 --profile uw4q9o7 jqfuczz 3cavmty --v 7
a long-lost VR headset worshiped as an idol by forest dwellers, adorned with feathers and vines, moonlight casting divine aura --ar 16:9 --profile uw4q9o7 jqfuczz 3cavmty --v 7
a towering futuristic city crumbling into primitive ruins, gleaming skyscrapers overtaken by wild jungles, neon lights fading into torchlight, flying cars replaced by tribal totems --ar 16:9 --profile uw4q9o7 jqfuczz 3cavmty --v 7 
a classroom of evolved AI robots devolving into cave dwellers around a fire, glowing embers illuminating their regression, wall carvings echoing ancient storytelling --ar 16:9 --profile uw4q9o7 jqfuczz 3cavmty --v 7 
a massive DNA helix unraveling backward, each rung turning into fossilized remains, floating in a dark void with subtle bioluminescence --ar 16:9 --profile uw4q9o7 jqfuczz 3cavmty --v 7 
cyborgs dismantling themselves into stone tools and bones, surrounded by fading holographic projections of past glory, misty atmosphere with dramatic lighting --ar 16:9 --profile uw4q9o7 jqfuczz 3cavmty --v 7 
Recent articles
Related articles
Free Geeky Animals' Sref code: 1319887298
Prompts are created using MJ Simple Prompt Generator (+ my brief idea)
Cover prompt: middle-aged woman in traditional attire selling fish, coastal market scene with sea breeze, fresh seafood on display, early Sunday morning, birds flying, rustic stalls, vibrant lighting --ar 16:9 --sref 1319887298 --profile i6fyyx5 --sv 4 --v 7
I hope you like this article!
Thank you for reading and happy creating!