Chase Lean Profile picture
Oct 27 10 tweets 4 min read Twitter logo Read on Twitter
It's hard to make consistent characters in DALL-E 3.

Until now.

By combining everything we know so far, it's time to solve this problem.

Here are the top 5 ways to create consistent characters with DALL-E 3:
Image
Image
Method 1:

The most straightforward way is to ask DALL-E 3 to generate multiple poses of the same character in a single image.

Example prompt:

a character sheet of a [chef] showing the same person from multiple angles
a character sheet of a chef showing the same person from multiple angles
Professionally designed vector illustration on a pristine white backdrop, presenting a comprehensive character study of a European male chef in his 40s from multiple angles. The sheet should depict him chopping vegetables, tasting a soup, kneading dough, and presenting a finished dish.
Instead of a character sheet, you can also use prompts like:

- concept art of a [person] from multiple angles
- comic strip depicting a [person] with multiple panels

You get the idea. Try to make many pics of a character in the same image.
Black and white technical drawing concept art on a white background, illustrating a character sheet of a medieval mage. The sheet showcases him from multiple angles: front, side, and back view. The mage is shown conjuring a spell, with swirling magic around his hands, studying an ancient tome, and channeling energy through a staff.
A comic strip in which a boy discovers a mysterious egg that hatches into a baby dinosaur.
This method has one limitation:

What if you want to generate consistent characters across multiple different images?

@anukaakash, @umesh_ai, @artichokes_art and @ai_for_success have recently shared excellent tutorials on how to do so.

Here are their discoveries:
Method 2 by @anukaakash

Using DALL-E 3 in ChatGPT, directly ask it to create 4 images of the same person - but with a different age.

You can change something other than the age.

Method 3 by @umesh_ai

Describe your character in extreme detail - his age, eye color, hair length, shirt color, etc.

Ask GPT to use this description as a reference every time it writes a prompt for DALL-E 3.

Method 4 by @artichokes_art:

Start with a character sheet (Method 1).

Then, use GPT to slowly modify this image into any style you like.

Method 5 by @ai_for_success:

First, write a detailed prompt (Method 3).

Then, change the prompt ever so slightly, while keeping the body of the prompt and the seed constant.

And that's it!

It's great to see what we can achieve as a community.

Everyone mentioned above posts some really high quality AI stuff - not just on character consistency.

I'd highly recommend giving all of them a follow!
That's all for now.

Hopefully you found this post useful.

1. Follow me at @chaseleantj for more high quality guides on how to use DALL-E 3.

2. Like and repost the first post below to share it with others!

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Chase Lean

Chase Lean Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @chaseleantj

Oct 26
Previously, I showed you guys how to make vector illustrations with DALL-E 3.

I just found a much better way.

You can even control how artistic/professional the vectors look to create endless variations.

Here's how to do it: Image
First, ask ChatGPT this:

Can you generate me some flat vector illustrations on a white background of 4 different people?

The white background is very important.

It improves the quality of the vectors and makes them look cleaner.
Here are the results:

Not bad!

But what if we want to make the images more formal and professional? Image
Read 9 tweets
Oct 21
Most people aren't aware of DALL-E 3's full potential.

I've spent the past 3 weeks deep diving into DALL-E 3 to uncover its capabilities.

Here are my top 10 tutorials:
2. How to make incredibly detailed drawings
Read 12 tweets
Oct 20
Just tried @morphaistudio, a new free AI video generator.

Unlike Pika Labs & Runway, they offer a default 1080p resolution with no limitations & can make longer videos (up to 7 sec) in one generation.

I did some tests, and the quality is pretty good.

Here’s how to access it:
I’ve partnered with @morphaistudio to bring you this post.

How to sign up for a free account:

1. Visit the Morph Studio website (link in comments).

2. Click the “Join Our Discord” button.

3. Accept their Discord invite.
When you’re in Discord, look for the Morph bot.

To do so, enter one of the video generation channels. Image
Read 10 tweets
Oct 18
Previously, I thought DALL-E 3 couldn't generate images with lots of detail.

Not anymore.

I've just found a way to make highly detailed illustrations with it.

Here are 7 incredible examples + the prompts I used: Image
1. Dragon boss fight

Note that all the prompts I used are in the ALT of the images. Black and white technical drawing concept art, illustrating a ground combat scenario between a nimble ninja and a massive dragon. The environment shows the aftermath of their battle, with scorched trees and shattered rocks.
2. Angel vs demon Black and white technical drawing concept art depicting a ground combat scenario where the white angel uses a shield of light to deflect the black demon's fiery projectiles. The environment around them shows signs of their intense battle, with scorched earth and shattered structures.
Read 9 tweets
Oct 14
The biggest challenge most people face when using Midjourney?

It’s hard to find high quality prompts.

But that’s changed.

@getremixai just launched an app where you can browse thousands of AI images, each paired with a prompt, all in one place.

Here’s how to access it:
I partnered with @getremixai team to bring you this post.

First, head over to the Play Store/iOS App Store.

Search for “Remix AI Images and Video” and install it. Image
It’s pretty straightforward to use the app.

Just swipe up and down to view the images in your feed.

Tip: Tap on the magnifying glass at the top right. Browse “Trending remixes” & “Most liked posts” to learn from the best.
Read 7 tweets
Oct 11
A few hours ago, Adobe released Firefly 2, their new AI image generator.

But how good is it?

I compared it to Midjourney & DALL-E 3 across 10 very different image categories so you don't have to.

Here are the results:
1/10 Realistic photos

Firefly 2 is super good at generating photos.

The shots are more realistic and have higher resolution than MJ, but both are good.

Meanwhile, DALL-E 3 lags behind. Image
2/10 Product photos

I'm pretty satisfied with all three.

I might prefer the Firefly one a bit more, because of the higher contrast making the perfume bottle stand out. Image
Read 12 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us on Twitter!

:(