Greeting Cards - Workflow Wednesday

[silver lily]

This week, we’re exploring GPT-4o Image Generation’s ability to create beautiful, personalized greeting cards! From nostalgic vintage prints to futuristic cityscapes, GPT-4o can generate stunning cards that reflect any style, era, or vibe—all from a simple prompt.

A few quick notes on where to post:

• Use this gallery for long-form, step-by-step workflow guides.

• Use https://discord.com/channels/974519864045756446/979617962728226826 for quick, one-shot creations.

For this theme, I decided to make "greeting cards through the ages", with a focus on the beautiful city of Athens Greece.

• 📜 Step 1: Ancient Greece
  I started with a vintage-style card featuring a philosopher and the Acropolis in the background—earthy tones, classical style, and “Greetings from Athens” in elegant script.

• ⚔️ Step 2: The Middle Ages
  I asked GPT-4o to evolve the image into a medieval version, keeping the Acropolis intact but shifting the aesthetic to highlight the Mediterranean Sea and a fortified city.

• 📷 Step 3: Early 1900s
  Then I requested a black-and-white postcard reflecting the invention of photography. The Parthenon still anchors the image, now surrounded by the growing modern city.

• 🛰️ Step 4: Modern Day
  I prompted for a drone-shot style card from 2024, with the Acropolis framed by Athens’ contemporary sprawl and the sea in the background—sunny, vibrant, and realistic.

• 🌆 Step 5: The Year 2192
  Finally, I jumped into the future! A neon cyberpunk skyline, flying cars, and towering skyscrapers now surround the Acropolis, which I made sure remained untouched. I updated the text to a futuristic font for that extra sci-fi flair.

With the power of ChatGPT, I didn't have to upload any images in the chat. Instead, I simply described how I wanted the previous one to evolve. This kept the framing, composition, and style consistent while letting each time period shine.

[fluid thicket]

Minimalist Mokuhanga Greeting Cards: I took advantage of 4o's ability to use visual references to create new images. I uploaded and discussed a set of minimalist mokuhanga images that I created a while ago with DALL-E 3.

Then, I told ChatGPT that I wanted to create a set of greeting cards in this style, and asked for a style guide for this project. Once we finetuned the style guide, I generated the variations by specifying: the occasion or mood, the cat's pose, and the base color.

I decided on Japanese logos on the whim, and was pleasantly surprised that 4o is able to correctly spell out Japanese messages in hiragana, katakana and kanji. Messages read, from the first to the last, "What a joyous occasion", "Pun-pun [onomatopoeia for anger]", "Take good care", and "Thank you".

[fluid thicket]

Conversion without Reference Images - In my workflow above, 4o had the visual references to recreate the style I wanted in a greeting-card format. Is it possible to transfer the style from the previsously generated image into a greeting-card format without the sample images? The workflow I've been developing involves an extensive discussion and iteration to create a JSON-style prompt template, which allows me to customize the subject, scene, mood and other details, while maintaining the same art style.

The important first step is to identify an image/s that exemplifies the style you are looking for. This can be a sample image you wish to emulate or perhaps an image you generated previsously with a different AI model. Or, you might be experimenting with a unique style and finally gets an image you really like. In this use case, I was playing with Surrealist styles and arrived at the style I really liked (first image). An obvious option is to ask ChatGPT to turn the generated image into a greeting card. However, the iteration process strips saturation and vibrancy from the color palette, which gets worse with each iteration (second image).

An alternative that I've been experimenting with is to get ChatGPT to generate a customizable JSON-style prompt template with highly detailed style specifications. It took several iterations to get it just right, but once I got the right spec language, I was able to adjust the season and lighting and add the white frame and the logo, while recreating the same style in a fresh chat with no conversational context or visual references (third image). This alternative is quite time consuming, I found, but it's worth the effort if I'm producing a series of images in the same style.

Upon reviewing the third image, I can see the yellowish tint creeping in. I will need to address that drift in the next iteration of my prompt template.

[hazy notch]

I used GPT4 to invent a unique day to celebrate with a pop up greeting card. After choosing the topic, I did everything else in 4o: “Here’s a structured breakdown to help you create the pop-up card for National Left Sock Remembrance Day:

⸻

Front of the Card:
• Material: Use a collage of mismatched sock patterns or real fabric swatches in various colors/textures.
• Text:
“In memory of all those left behind…”
Use a slightly formal font with a wink of melodrama—almost like a vintage funeral notice, but cheeky.

Inside of the Card (Pop-Up Section):
• Pop-Up Element:
A single left sock (fabric-textured cardstock or felt) pops up from a small laundry basket or dryer door that opens slightly.
You could have the sock hang on a mini paper clothesline or float up with ghostly sparkles as if ascending to the “dryer dimension.”
• Tagline:
“Gone but not forgotten… to the dryer dimension.”
Add a little constellation of missing socks in the background, like a galactic memorial wall.
• Optional Pop-Up Enhancements:
• A faint “whoosh” line to show the sock vanishing.
• A dryer door that opens with a ghostly glow.
• A calendar marking April 23rd as National Left Sock Remembrance Day.

⸻

Back of the Card:
• Include a small dedication:
“This card is lovingly dedicated to the mysterious forces that claim our socks. May your left one find peace.”