Skip to main content
Trevor.Dennis
Community Expert
Community Expert
Nov 1, 2026 5:30 AM UTC
StickyQuestion

A Surprising Generative AI Result

 This amazed the heck out of me, so I have to share.  My wife is President of a local Garden Club and asked me to help create a poster for an event.  Because she has never used Generative AI, she was not encumbered with thoughts that limited what you might reasonably ask of it, so she gave me this completely (I thought) OTT prompt to paste in its entirety into the Prompt to Edit text field:

Create a premium A4 Christmas Lunch invitation for the Awatere Garden Club. Surround the entire invitation with a lush, realistic floral border dominated by bright red New Zealand pōhutukawa flowers, silver fern fronds, kawakawa leaves, mānuka foliage, native greenery and a few hydrangeas and roses. The border should be abundant and elegant with no empty corners. Add subtle gardening elements at the bottom including a vintage watering can, garden trowel and gardening gloves. Centre panel in warm cream parchment with sophisticated typography in deep green and Christmas red. High-quality print-ready design.

Text:

YOU'RE INVITED

AWATERE GARDEN CLUB

Christmas Lunch

Celebrating another wonderful year of gardening, friendship and community

Thursday 3 December

10.00 am

Meet at ** ********** Street, Springlands

11.15 am
Sales Table, Raffle & Chat

12.00 noon
Two-Course Christmas Lunch

Venue
Vintage Car Club
Brayshaw Park

Cost
$42 per member

Join fellow members for a festive Christmas gathering surrounded by friends who share a love of gardening. Enjoy delicious food, festive cheer, a raffle, a sales table, and plenty of time to chat and celebrate another successful gardening year.

RSVP by 27 November, 2026
*********@gmail.com

Merry Christmas & Happy Gardening!

AWATERE GARDEN CLUB

Style: realistic botanical illustration, elegant, professional, festive, New Zealand garden club, high-resolution printing, abundant pōhutukawa and native foliage.

I started with an A4 document and downsized it to 2000 pixels on the long side, and used GTP Image, which has been working well for me.  I totally expected it to start smoking and burst into flames, but this is what it gave me after about ten seconds.

 

    11 replies

    Participant
    Nov 18, 2026 10:07 AM UTC

    That's genuinely impressive! The level of detail in the prompt is quite ambitious, but the result shows how much generative AI has improved at understanding complex design instructions. The combination of native New Zealand flowers, festive elements, and elegant typography makes this a great example of AI being a practical creative assistant. Sometimes the best results come from giving AI a detailed brief and letting it surprise you. Thanks for sharing!

    Translate
    Chuck Uebele
    Community Expert
    Community Expert
    Nov 11, 2026 1:35 AM UTC

    Wonderful work! AI has come a long way.

    Translate
    justlearn
    Community Expert
    Community Expert
    Nov 5, 2026 11:59 PM UTC

    ​@Trevor.Dennis This is such an empressive turn-out! If you think about it, if you did this same prompt 2-3 years ao, you would have gotten a ‘hot mess’ — it’s amazing how AI technology improves...hopefully not to the point where ‘graphic designers’ are no longer needed! 

    Translate
    m
    Bruce Jones Harry
    Participating Frequently
    Nov 5, 2026 2:38 AM UTC

    I volunteer to be the official dessert inspector. Any objections? 😂

    Translate
    Gord@APL
    Inspiring
    Nov 2, 2026 4:53 PM UTC

    The quality/accuracy of the text and the detail in the flowers and leaves is phenomenal until you start zooming in… but there’s one three-fingered (or thumbless) gardening glove on top of another one with very odd proportions and angles to the fingers (or possibly two thumbs?). 

    Translate
    Trevor.Dennis
    Community Expert
    Community Expert
    Nov 1, 2026 8:10 PM UTC

    I tried the same prompt with Gemini and the result was less impressive.   I think it used 10 credits.

     

    Translate
    Community Expert
    Nov 1, 2026 7:17 PM UTC

    It’s impressive that text—which, until just a few months ago, was one of the biggest challenges for AI image generation—is now handled so well. Anyway, the prompt is very well-crafted, and the style is clearly defined. I wish businesses in my city knew how to do this—every few steps, I run into a sign that looks just like the last one. Or, actually, maybe it’s better if they don’t learn, and keep believing that you need graphic designers to produce creative, distinctive work.

    Translate
    Trevor.Dennis
    Community Expert
    Community Expert
    Nov 1, 2026 11:28 PM UTC

    I’ve just asked my wife and she tried to do it herself using copilot.  It asked her for content line by line (venue, date, cost etc.)  She told it she wanted nice NZ native plants and flowers, and created the prompt.  She was unable to proceed from there, and I have no experience with Copilot, so while I doubted it would work, it was worth risking ten credits to test it.

    I’ve used Claude with a Firefly project, but I can’t remember the details.  I remember it did exactly what I needed though.  Even using templates it would have taken a good hour to do that, and it would have been a big ask finding a template that was NZ native plants specific.

    Jesús Rameriz and Lisa Carney have repeatedly said what a huge time saver Generative AI is for them.  AI hold-outs will struggle to remain competitive against people who can undercut them.  They have to deliver artwork suitable for multiple formats, TV 16:9, cinema 2.4:1 (16:9 is 1.78:1), cell phone, Macbook 16:10 etc.  They use Generative Expand to fill borders.  

    Working from a blank screen can be problematic when you have a specific stock image in mind, so I use the AI a lot for fine tuning stock images.  It has not taken long for use to become quite dependent on the AI tools, and I would _really_ hate to have to go back to working without them.  That’s a real world point of view underlined by professionals like Lisa and Jesús so I hope Adobe does not give too much credence to the vocal minority who are afraid of change. 

    Translate
    jane-e
    Community Expert
    Community Expert
    Nov 1, 2026 5:50 PM UTC

    Very impressive Trevor!

    Jane

    Translate
    Community Expert
    Nov 1, 2026 4:15 PM UTC

    That's a seriously impressive result, Trevor, especially from a single prompt!

    One thing that caught my eye was the "Grow Nourish Bloom" text on the watering can. That's not in the original prompt, so I'm curious where it came from. Did the AI invent it, or is it possibly something from an existing product or image?

    It also raises a few questions for me from a design and prepress perspective.

    Do we know whether all the flowers, foliage and other elements are entirely AI-generated, or whether any existing imagery is being incorporated? And where would that leave licensing and commercial usage?

    Same with the typography. It looks great, but are those actual fonts or just rendered representations? What happens if you need to change the wording, adjust the typography or make a small layout correction?

    I'm also curious about the print side. You mentioned 2,000 pixels on the long edge, which would work out around 171 PPI at A4. Have you tried printing it, and how did the colours and smaller text reproduce? What about CMYK conversion and bleed?

    I'm not questioning how impressive the result is, because it really is. I'm just interested in how far it can go beyond generating a lovely-looking image and into producing something that's properly editable, reproducible and ready for print.

    For a one-off invitation, it could be absolutely ideal. I'm just curious about the practical limitations once you start treating it as finished artwork rather than a generated image.

    Translate
    Trevor.Dennis
    Community Expert
    Community Expert
    Nov 1, 2026 7:53 PM UTC

    Those are good points Eugene.  In this instance it is not going to be printed, but having seen what the AI model can do, and we know that we can specify colour codes, so I wonder if we could prompt to make it suitable for CMYK and prepress.  I am infamous for not caring about colour accuracy so long as it looks good but a lot of people would need the colours to be right.

    The pixel size would be easily overcome by upressing with Gigapixel.  I got a kick start on Adobe’s Generate AI models watching Lisa Carney’s presentation to the Adobe User Group of New Jersey about three years ago.  At that time we were more limited with image size, so Lisa worked at that 1024 pixel limit, and upresed the final image with Topaz Photo AI which had similar abilities as Gigapixel.  I got a Photo AI licence that same day and I have Topaz Studio now, and the full version of Gigapixel has more options than the Photoshop version. The text looks over crisp to me but I imagine this would print OK.

    4X

    6X (8484 X 12000)

     

    Regards editing the text, it would take a few minutes to match the font, which is a fraction wider than most, but Calisto MT is close.

    But the quick way, if you have the credits is to instruct Flux to do it.

    Translate