Great NEWS! The wait is finally over! Today, Midjourney is testing a new version of the /describe command for V6.
With/describe, it allows you to go from image to text. It’s an Image-to-Text Command! This enhanced version generates longer and more intricate prompts tailored specifically for V6.
It works even better when combined with Midjourney latest feature "Style References". Today, let me show you how to replicate an image easily.🎨
Have you ever seen a mesmerizing work of art and thought — I wish I could recreate something just like this! Well, now you can with Midjourney AI and some prompt engineering.
Today, we will use the new “/describe” command and “Style Reference” to replicate any images you want!
I’ll walk you through simple step-by-step instructions for replicating any images like this below.
What is /describe on Midjourney?
⚒️ Function: Midjourney offers the /describe command to convert images into text prompts. It creates 4 text prompts based on an image.
⚙️ Process: The tool reverse-engineers the uploaded image to generate a set of descriptive prompts.
🖼️ Prompt Generation: Descriptions are created using inspirational and suggestive words, maintaining the aspect ratio of the original image.
🎨 Image Creation: Users can use the suggested prompts to create new AI-generated images while preserving the aspect ratio.
📐 Aspect Ratio: Midjourney preserves the aspect ratio of the uploaded image when generating new images.
🔥 The NEW /descibe will produce longer and more detailed prompts specifically tuned for v6
What to know
Conversion: The Describe tool on Midjourney converts images to text prompts.
Recreation: Users can recreate original images on Midjourney using the AI-generated prompts.
Today’s content will be broken down into two sections:
Section One: Simple Guide to use “/descirbe”
Section Two: Combine the Power of “/describe” and “Style Reference”
Let’s start with the basics!
Part One: Simple Guide to use “/describe” 🎨
Step 1: Type the command /describe
Type the command /describe and you will be given two options: Image and Link. You can either upload an image or paste an image link that you found on the internet.
Step 2: Four Possible Prompts
Click “enter”. Midjourney will analyze your uploaded image and suggest four prompt ideas for you.
Step 3: Choose the preferred prompt
Users can choose from the suggested prompts or combine them to visualize new images.
You can then generate images for each prompt by clicking on the relevant 1️⃣2️⃣3️⃣4️⃣ buttons underneath or select “Imagine All” to generate all four variations at a time.
4 The Results
Here are the four variations generated with the NEW “/describe”. I will say it’s pretty good and I love the aesthetics
Limitations of /describe
Note that the “Describe” tool may not produce images that are identical to the original everytime, but
📌 Tip: You may want to edit the prompts to refine the AI-created images.
How to edit
To edit the prompt, select one of the four prompts generated with /describe.
You will then be given a text box to make your edits.
Section Two: Combine the Power of “Describe” and “Style Reference” 🎨
Besides customizing the suggested prompt, you can also use “Style References” to prompt a more accurate replication
Now, let’s see how you can leverage the power of “Style Reference” to level up the replication!
Type/imagineas you usually do. Then, copy and paste the text prompt generated with the “/describe” command.
We are using the suggested prompts from above.
To use “Style Reference”, simply type --sref after your text prompt, followed by URLs to the same image you uploaded when using the/describe command
The Results
Here is the result of combining the text prompt generated with the new “describe” and “Style Reference”
Guide to Style Reference
👉🏻 If you are not familiar with Style Reference, it is the latest feature of V6 that allow you to generate a new image that visually ‘matches’ the aesthetics of a reference image.
I have written a Thorough guide with 4 Advanced Tips
Final Thoughts
Although the team is still fine-tuning the details, this enhanced version of “describe” aims to generate longer and more intricate prompts tailored specifically for V6.
I highly suggest using “describe” together with “Style Reference” if you are looking for more accurate image replication.
Comments