Bring household objects to life with two simple GPT4 prompts!

Looking for inspiration for digital art?

Take a photo of a household object and bring it into the world of DALL-E 3.

Overview

Upload your image to Chat GPT (default, not DALL-E)

• Upload your image into ChatGPT (default, not DALL-E)
• Prompt ChatGPT to reply with a description you can use to create a similar image in DALL-E.
• Start a new conversation in ChatGPT (DALL-E 3 option), provide the prompt

Step-by-Step Process

Upload Your Image to Chat GPT

Start a new conversation with ChatGPT.

Make sure “Default” is selected in the “GPT-4” drop-down. We will use the “DALL-E” option later, but DALL-E doesn’t allow you to attach images.

Click the image icon on the left side of the field where you type messages to GPT. Attach the image you want to work on.

Don’t send the message to ChatGPT yet.

Include this Prompt with the Image

In the same message as the attached image, send this prompt:

This is a photo of a sculpture of a deer made from parts of a birch tree. Exclude any mention of the pottery or the blue vase. please reply with a MECE description i can send to DALL-E to prompt the creation of an image as similar as possible to the deer sculpture in this photo. Make the description a 5 level deep bullet point list.

Of course, you’ll revise the prompt to describe your photo.

You’ll notice that I provide some hints to DALL-E. DALL-E has a limited amount of attention for each interaction. If you point it towards the items and details you want to focus on, you’ll get better results.

GPT Will Reply with a MECE Description of Your Image

MECE = mutually exclusive, collectively exhaustive.

Mutually exclusive means “don’t repeat anything.”

Collectively exhaustive means “include all the necessary info”.

GPT understands MECE, so you can use it in any prompt without defining it.

Here’s GPT’s reply to this prompt:

1. Overall Structure
   • Head:
     • Cylindrical birch section
     • Positioned looking forward
     • Proportionate to body
   • Body:
     • Elongated cylindrical birch section
     • Horizontal orientation
     • Natural bark texture
   • Legs:
     • Four individual birch branches
     • Positioned symmetrically (two on each side)
     • Upright stance
2. Antlers
   • Originating from top of head
   • Made of thin birch branches
   • Two main branches:
     • Extending upwards and outwards
     • Multiple smaller branches stemming off main branches
3. Facial Features
   • Eyes:
     • Subtle indentations
     • Positioned symmetrically on the head
   • Ears:
     • Two smaller birch sections
     • Placed behind and above eyes
     • Tilted slightly backwards
   • Nose and Mouth:
     • Not distinctly carved or differentiated
4. Texture and Coloring
   • Natural birch bark texture:
     • White with black horizontal lines and markings
     • Rough and uneven surface
   • No added coloring or paint:
     • Authentic wood appearance
     • No alterations from natural birch look
5. Size and Proportion
   • Life-sized:
     • Comparable to a medium-sized deer
     • Legs proportionate to body
   • Balanced structure:
     • Even weight distribution
     • Stable positioning on flat surface

Added Bonus

An important added bonus of this MECE description is that it gives you a sort of form you can use to change any aspect of the image DALL-E will produce, with minimal disturbance to the rest of the image.

Copy / Paste the MECE Description from GPT to GPT/DALL-E

Start a new conversation in GPT4.

Choose “DALL-E 3” from the list of options for GPT4. This tells GPT that you want it to create an image for you.

Paste in the MECE description you copied from your chat with GPT4.

Send this message and GPT4/DALL-E will reply with an image that we hope is very close to the image you provided. So far I’ve been very impressed with the results.

Here’s the image I got from my MECE prompt:

Chat with GPT4/DALL-E to Revise the Image

Once you’ve got your real-life image imported into DALL-E, you can chat with DALL-E to change it any way you want. Add a background scene. Change the figure.

Have fun!