Creativity AI #51: Midjourney V8 teased, Claude Opus 4.5 drops, and AI wants to do your shopping
Office hours updates, productivity research that shows the math, and a $1 ElevenLabs deal
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 Office Hours (26th November 2025): Style Creator upgrades, V8 inches closer
David was 15 minutes late (classic), but the office hours packed in some solid updates.
Style Creator improvements
The team wants to make Style Creator more powerful. Better color control and metaphor handling are on the way. They’re also testing a scrolling interface instead of the current grid, which could help users move toward what they want more intuitively.
Improvements could land in a week or two.
One technical detail worth noting: codes generated by Style Creator are brand-new codes, not random sref codes. And here’s the key difference between stacking sref codes versus using Style Creator: stacking finds an average between style points, while Style Creator finds the exact point you’re after. That matters if you’re precise about your aesthetic.
V8 is coming (but not for Christmas)
The developers are getting closer to V8 training. No holiday release though. V8 will be undergoing training during Christmas, with a possible early 2026 release for two models: V8 and another new editing model.
Expected V8 features:
Better text generation
Improved coherence and knowledge
A new Edit model that takes multiple reference images (clothes, objects, etc.) to build scenes and stories
New retexture capabilities
Fair warning: V8 will require learning new prompting approaches.
V7.1 might get scrapped
V7.1 also requires new prompting methods, and old prompts may not work. Real-time preview works well, but the team seems hesitant. Unless there’s a killer feature (currently there isn’t), they might skip straight to V8.
Other things
David reiterated that short prompts work better. LLM-generated long prompts aren’t optimal.
95% of images now use V7. Older models account for just 5%, with Niji being the most popular among those.
If Niji gets an update, it’ll be Niji 7, not 8.
Midjourney plans to hire more people. Lots of ideas, not enough hands (3D features are on the wishlist).
The underlying Sref system needs to be fixed after V8 launches (current issues include blurry images).
David called Midjourney a “vehicle for imagination,” but clarified they’re not building self-driving cars.
Claude Opus 4.5 arrives, and it’s actually affordable this time
Anthropic just dropped Claude Opus 4.5, calling it “the best model in the world for coding, agents, and computer use.” Bold claim. But here’s what caught my attention: the pricing.
At $5/$25 per million tokens, this isn’t the wallet-draining Opus experience we’ve had before. Previous Opus models would burn through my usage in a day. Literally one day. I’m skeptical but hopeful this time. Check back with me in a week.
The performance numbers are wild, though. On Anthropic’s internal engineering exam, Opus 4.5 scored higher than any human candidate within the 2-hour time limit. It handles 30-minute autonomous coding sessions without falling apart. And it uses 48-76% fewer tokens than previous models while delivering better results.
Some standout capabilities:
Elite software engineering, especially complex multi-system bugs
Long-horizon agentic tasks with fewer dead-ends
Best-in-class prompt injection resistance (harder to trick than any other frontier model)
Actually understands what you want on the first try
Partners are reporting 50-75% fewer errors and up to a 65% reduction in tokens on complex tasks. Multiple testers say it has “the best frontier task planning and tool calling we’ve seen yet.”
The key shift: tasks that were “near-impossible for Sonnet 4.5 just a few weeks ago are now within reach.” If you’ve been avoiding Opus because of cost, this might finally be the version worth trying daily.
Link to the video
Claude says AI boosts productivity by 80%, and shows its math
Anthropic published research on AI productivity gains, and the methodology is interesting. They sampled 100,000 real Claude.ai conversations, then used Claude itself to estimate how long tasks would take with and without AI help.
The headline number: 80-84% median time reduction on tasks.
Some specific examples:
Curriculum development: 4.5 hours → 11 minutes
Writing documents and memos: 87% time savings
Compiling information from reports: ~95% savings
Checking diagnostic images: Only 20% savings
The economy-wide projection suggests a potential 1.8% annual increase in US labor productivity over the next decade. That would nearly double recent growth rates.
Software developers account for 19% of total AI productivity gains. General managers, market research analysts, and customer service reps follow behind.
One important caveat: these estimates don’t account for time spent verifying AI outputs. Real-world trials found smaller savings (14-56%) because users still need to check and polish what AI produces. Physical tasks and real-time judgment calls see minimal gains.
Still, if your work involves writing, research, or synthesizing information, the potential speedup is significant.
OpenAI and Perplexity launch AI shopping assistants
Both companies announced shopping features this week. OpenAI partnered with Shopify for in-chat checkout, while Perplexity teamed up with PayPal.
Users can ask for product recommendations with specific criteria, share images of items they like, and complete purchases without leaving the chat. Adobe predicts AI-assisted shopping will grow 520% this holiday season, so the timing makes sense.
I see the appeal for quick purchases. But for anything requiring real consideration, I’m not convinced a chatbot beats actually browsing and comparing yourself. What if I miss out on a discount code or voucher on the website?
ElevenLabs Black Friday deal: Starter Plan for $1
ElevenLabs is running a Black Friday sale with their Starter Plan at just $1. You can use it to create AI audio, images, and video. No code needed, just sign up and start creating here.
Update from Geeky Curiosity
Five free Midjourney fast hours just for setting up a Midjourney Profile page. I grabbed mine last week. Hope you did too!
The whole thing took maybe two minutes. Midjourney’s offer only lasted 24 hours after the Profile feature launched, so if you missed it, that window’s closed.
Follow along here so you don’t miss the next one. I’ll share the latest updates (esp. when there’s free or discount things) from Midjourney and other major AI companies.
Refreshed Midjourney Style Resources
The Midjourney Style Resources #30 has been updated to V7 aesthetics.
The samples from Resource #30
Geeky’s music
Link to music
Inspiration
Soho Rep’s theatre posters prove that creative freedom within structure produces extraordinary work. Their process blends collaboration between illustrators, playwrights, and directors with unapologetically bold visual experimentation.
The magic happens in the balance. Creative directors provide detailed briefs with sketched concepts and inspirational imagery, then step back. Illustrators dive deep into play texts, exploring language nuances and interpreting beyond strict guidelines. This freedom allows artists to deliver ideas that surpass initial expectations.
Embrace the weird. Soho Rep’s posters don’t aim to be universally “pleasing”—they’re bold, unconventional, sometimes uncomfortable. The only requirement is to reflect the spirit of the work. The aesthetic diversity matters.
Yet, don’t expect Midjourney to get the text right, though.
Bold manifesto poster that leads with language, giant unapologetic letterforms carved like architectural blocks across a calm field of negative space, eccentric ligatures colliding with minimalist grids, weird and distinctive, eye-catching color planes that slice diagonally, small typographic footnotes revealing nuanced intent, experimental texture from print grain and edge bleed, visual storytelling that feels like a rallying cry for freedom, personality, and flair, the words themselves becoming extraordinary sculpture, inviting an immersive read from far away to nose-close detail --ar 2:3 --profile ppjylid --v 7 
Minimalist poster with a single weird icon, nuanced geometric creature built from simple shapes, generous negative space, crisp micro-typography whispering context, experimental color pop on matte field, distinctive and extraordinary, quiet yet bold --ar 2:3 --profile ppjylid --v 7 
Eclectic Specimen Wall poster, diverse illustrative artistry arranged as a clean grid, each tile a different style, minimalist captions, fresh lens on taxonomy, freedom and flair balanced by ruthless negative space, unconventional yet cohesive visual storytelling --ar 2:3 --profile ppjylid --v 7 
An unapologetic print experiment channels Riso misregistration, bold shapes and halftone fields wander slightly out of alignment to create living color, grainy textures and overprint halos hum with energy, language-focused labels act as anchors, minimalist grids resist the chaos just enough, a weird and distinctive artifact that celebrates process, freedom, and surprise, immersive storytelling in the edges and overlaps, a fresh lens that turns production quirks into real art with extraordinary personality and flair --ar 2:3 --profile ppjylid --v 7 
Style & Approach
bold, weird, unapologetic, eclectic, distinctive, eye-catching, experimental, unconventional, immersive, collaborativeVisual Techniques
intricate illustrations, minimalist graphics, diverse illustrative artistry, language-focused design, negative space, aesthetic experimentation, visual storytellingCreative Characteristics
freedom, personality, flair, extraordinary, nuanced, fresh lens, deep dive, surprising, real artProcess & Methodology
collaborative, multi-disciplinary, interpretive, brainstorming, workshopped, detailed brief, inspirational imagery, in-person exchangeRecent articles
Related articles
Free Geeky Animals' Sref code: 3293057002
Prompts are created using MJ Narrative Prompt Generator (+ my brief idea)
. --sref 3293057002 --profile ppjylid --v 7 
Cavers descend into quartz cathedral, walls singing with low resonance, bio-lights trapped in facets, rope lines taut, headlamps carving shafts, sacred hush of discovery, crystalline dust glittering in air --ar 16:9 --sref 3293057002 --profile ppjylid --v 7 
Skiff skimming methane sea, oars barely kissing mirror surface, drifting spore-rafts glowing amber, storm front brewing turquoise, horizon split by lightning, exploratory calm before revelation, contemplative pacing, Malick-like reverence --ar 16:9 --sref 3293057002 --profile ppjylid --v 7
Three scientists circle humming monoliths, glyphs pulsing like breaths, twin suns bleaching the plateau, dust banners in crosswind, laser scanner painting lines, tense wonder at first contact, precise framing, stark color separation --ar 16:9 --sref 3293057002 --profile ppjylid --v 7 
Cover prompt: Bold manifesto poster illustration that leads with language, giant unapologetic letterforms carved like architectural blocks across a calm field of negative space, eccentric ligatures colliding with minimalist grids, weird and distinctive, eye-catching color planes that slice diagonally, small typographic footnotes revealing nuanced intent, experimental texture from print grain and edge bleed, visual storytelling that feels like a rallying cry for freedom, personality, and flair, the words themselves becoming extraordinary sculpture, inviting an immersive read from far away to nose-close detail --ar 16:9 --profile ppjylid --V 7
I hope you like this article!
Thank you for reading and happy creating!
Wow, the part about Style Creator finding the exact point versus sref stacking just averaging styles really stood out; I'm curious if that implies Style Creator navigates a more determinictic path within the latent space.