Step by Step Midjourney Prompts Generation using ChatGPT 4o’s Vision Analysis

Rifx.Online
Natural Language Processing , Generative AI , Technology/Web
15 Dec, 2024

You know that feeling when you’re trying to create the perfect Midjourney prompt, but the words just don’t capture what’s in your mind? I’ve been there too. After countless hours of experimentation, I’ve discovered a game-changing approach: using ChatGPT 4o’s vision capabilities to analyze reference images and generate precise prompts.

Let me share my journey and the exact process I use. Trust me, this will transform how you create AI art.

The Lightbulb Moment 💡

I used to spend hours describing what I wanted to Midjourney. Sometimes it worked, often it didn’t. Then it hit me — why describe when I can show?

ChatGPT 4o’s vision analysis became my secret weapon. Instead of struggling with descriptions, I now let it analyze images I love and translate them into perfect prompts.

Getting Started: The Image Collection Phase

First things first — gather your inspiration. But here’s a crucial tip I learned the hard way: don’t go overboard with images.

From my experiments, the sweet spot is no more than 25 images total per conversation. This keeps everything manageable and ensures detailed analysis.

Here’s the game-changing insight I wish I knew when I started: the quality and diversity of your collected images directly impact your final results. Think of it like teaching by example — if you feed ChatGPT diverse, high-quality references that truly represent what you want, it will understand your vision much better. I’ve seen massive differences in prompt quality just by being more selective with my reference images.

I break this into batches of 10 images each. This systematic approach has consistently given me the best results.

The Magic Formula: First Analysis Phase

Here’s the exact prompt I use for the initial analysis (feel free to copy it):

Please analyze the uploaded images in detail.
For each image, identify:
Key elements (e.g., objects, patterns, and
recurring motifs).
The artistic style used (e.g., paper-cut,
abstract, minimalist).
The dominant color palette and transitions
between shades.
The theme or mood conveyed by the
composition.
Provide your findings in a format similar to
this:
Image 1 (Full Moon with Cherry Blossoms)
Elements: [description]
Style: [description]
Color Palette: [description]
Theme: [description]

Why This Works So Well

Think of it like having a professional art critic analyzing your favorite pieces. The structured format helps ChatGPT break down each image systematically.

I’ve found this approach captures details I might have missed myself. It’s like having an extra pair of eyes that sees everything.

The Two-Batch Strategy

Here’s my time-tested approach:

**Batch 1 (First 10 Images)**Upload your first set and use the analysis prompt. Take a moment to review the findings.

**Batch 2 (Next 10 Images)**Repeat the process with your second batch. This builds a comprehensive understanding of your style preferences.

Finding Patterns: The Secret Sauce 🎯

After analyzing both batches, I use this simple but powerful prompt:

based on details in all uploaded images in
this conversation and Identify common
patterns and predict top 5 most preferred
objects and design patterns

This is where the magic happens. ChatGPT starts connecting the dots between all your references.

Crafting the Perfect Prompts

Now comes the fun part. Here’s my final prompt that consistently delivers amazing results:

using the preferred objects and design
patterns craft 6 Midjourney AI prompts that
will create a series of close-up vector
images that illustrate 3D red paper cut for
Chinese New Year and parameters --ar 16:9
--stylize 50 --v 6.1  
exclude dragon and tiger as part of the
prompt
and following are prompts example, can
analyze them to understand the Midjourney AI
prompt structure better, but not copying
their details: 

A red paper-cut of a peony flower. It is
simple and pretty. --ar 16:9 --stylize 50 --v
6.1
A beautiful paper-cut card with soft colors.
It shows a child happily playing with an
abacus at the Mid-Autumn Festival night.
There are lanterns and a family together. It
is mostly red. --ar 16:9 --stylize 50 --v 6.1
A red paper-cut of a peony flower. It is
simple and pretty. --ar 16:9 --stylize 50 --v
6.1
A flat paper-cut of a pretty Chinese window
flower. The flower has simple shapes and
smooth lines. It is on a white background.
--ar 16:9 --stylize 50 --v 6.1

The Structure That Works

Every successful prompt follows this pattern:

[Main Subject] + [Style Description] + [Additional Details] + [Technical Parameters]

Think of it as a recipe — each component plays a crucial role in the final result.

Avoiding Common Pitfalls

Through my experiments, I’ve learned to watch out for:- Overcomplicated descriptions- Missing technical parameters- Mixing conflicting styles- Unclear perspectives

Keep it simple and focused. Less is often more with Midjourney prompts.

A beautiful 3D red paper-cut featuring a
pagoda at the center, surrounded by swirling
clouds and symmetrical cherry blossoms. The
design showcases a layered composition with a
harmonious balance of traditional Chinese
elements --ar 16:9 --stylize 50 --v 6.1

A close-up of a peony flower in a 3D red
paper-cut style. The layers highlight the
petals' intricate patterns with fine details
on each leaf. The background includes subtle
curves and smooth circular shapes, enhancing
the elegant simplicity of the composition.
--ar 16:9 --stylize 50 --v 6.1

The Beauty of Simple Prompts 🎯

Here’s something fascinating I discovered during my experiments: Once ChatGPT understands the context of Midjourney prompt creation through our earlier interactions, we can switch to incredibly simple prompts. For example, I now just ask “describe the attached images using English that 7 year old can understand” and get perfectly detailed, clear descriptions. ChatGPT 4o’s memory of our previous Midjourney discussions helps it provide descriptions that work beautifully for prompt generation. This simplicity is powerful — it makes the whole process more approachable and less technical. It’s like having a conversation with a friend who already knows what you’re trying to achieve.

A red paper-cut of a pagoda surrounded by
clouds and cherry blossoms. The intricate
layers emphasize symmetry, with delicate 3D
cutouts showing fine details on the pagoda's
roof and branches. The background features
flowing organic shapes and swirling clouds.
--ar 16:9 --stylize 50 --v 6.1

An artistic 3D red paper-cut depiction of
plum blossoms on curved branches. The design
showcases organic flow with smooth layers and
subtle shadowing, embodying elegance and
natural beauty. --ar 16:9 --stylize 50 
--v 6.1

A 3D red paper-cut design featuring a
blooming peony surrounded by intricate
flowing branches and delicate foliage,
symbolizing prosperity and beauty. The design
is highly detailed with layered cuts to
create depth. --ar 16:9 --stylize 50 --v 6.1

The Power of Versatility 🎨

Here’s what I love most about this approach: it’s incredibly flexible. Just change your set of sample images, and you’ll get an entirely new set of prompts tailored to your new direction. Want to switch from Chinese paper-cut style to cyberpunk aesthetics? No problem. Just feed ChatGPT a new set of cyberpunk references, and it will generate fresh prompts matching that style. This makes it a universal prompt-generation framework that adapts to any artistic style you’re interested in exploring.

A 3D red paper-cut featuring a traditional
Chinese pagoda surrounded by pine trees and
layered clouds. The pagoda’s intricate roof
design is highlighted against a red
background with subtle white accents for
contrast. --ar 16:9 --stylize 50 --v 6.1

An elegant 3D paper-cut composition in rich
red with pure white underlays, showing ornate
pavilions nestled among cloud patterns.
Architectural details reveal multiple depth
layers, with white accents highlighting roof
tiles and lattice work. Intricate geometric
borders frame the scene with dimensional
shadows. --ar 16:9 --stylize 50 --v 6.1

A close-up of a serene full moon surrounded
by delicate cherry blossom branches in a 3D
red paper-cut style. The blossoms are
detailed with layered petals, and the
background is textured to emphasize depth.
The composition exudes elegance and
tranquility. --ar 16:9 --stylize 50 --v 6.1

A flat 3D red paper-cut illustration of a
blooming magnolia branch with smooth petals
and curved leaves. The design includes
minimal background swirls, focusing on
simplicity and elegance. --ar 16:9 
--stylize 50 --v 6.1

A delicate 3D red paper-cut of a full moon
surrounded by cherry blossom branches and
swirling clouds. The design is framed with
layered circular patterns, creating an
elegant and peaceful atmosphere --ar 16:9
--stylize 50 --v 6.1

Final Thoughts 🌟

This method has revolutionized how I create with Midjourney. It’s not just about generating prompts — it’s about understanding your artistic preferences and translating them effectively.

Remember, the key is patience and systematic approach. Don’t rush through the analysis phase — those insights are gold for creating perfect prompts.