Poe Ai Image Generators - 'Stable Diffusion 3' + 'Flux Pro'

Let’s explore the basics of AI image generation using the Poe Ai platform, which I’ve found to be particularly versatile and user-friendly.

Before we start, check out my article on How to Type Prompt with Poe Ai.

Poe’s platform offers it’s users a variety of Ai image generators, including ‘cutting-edge’ models like Stable Diffusion 3 and FLUX Pro.

Here’s an over-simplification of what it’s like to use an Ai image-generator on Poe, or anywhere else:

Prompt Input: You provide a text description of the image you want to create.
AI Processing: The Ai interprets your text prompt and does a lot of computing.
Output: 20 seconds later, you get unique image, based on your description.
Refinement: You can improve the image by modifying your prompt

Poe Ai Image Prompt Example — This is a Poe Ai chat window, showing the text-prompt in purple, and the resulting image is beneath, made by Stable Diffusion XL,

Comparing Image Generators in Poe Ai

Poe offers access to various Ai image generator models, each with its own strengths.

I gave the following text prompt to each of Poe’s 10 “Official” image generators:

“Close-up portrait of a reddish-brown Boxer dog, expressive eyes, wrinkled brow, soft focus background, collar visible, photorealistic, high-quality, detailed”

Below, you’ll see the images that resulted from that prompt, as interpreted by the various Ai image generators in Poe.

DALL-E 3

Excellent for illustrations but struggles with photorealism.

The dog looks like it was hand illustrated, and that’s a good look for certain applications.

For the purposes of creating a realistic-looking dog image, DALL-E-3 ranks at the bottom of the list of 10.

DALL-E-3 Example Image Output from Poe Ai — It took about 18 seconds to make an image in DALL-E-3.

Stable Diffusion 3

My current go-to for high realism and prompt accuracy.

I love that the dog image it made had a slight angle to it’s gaze, and one of it’s eyes is clearly in shadow.

Realistic aspects seem to be where Stable Diffusion 3 excels.

Stable Diffusion 3 Example Image Output from Poe Ai — It took about 5 to 8 seconds to make an image in Stable Diffusion 3.

Stable Diffusion 3 Turbo

Really-fast and dog images look good.

It’s evident that the images output are not as detailed when compared to another image-generator’s, such as regular Stable Diffusion 3.

The speed at which these images are created might be a factor for you, so if you want a fast image, Stable Diffusion 3 Turbo seems perfect.

Stable Diffusion 3 Turbo Example Image Output from Poe Ai — Stable Diffusion 3 Turbo made images in about 1.5 seconds.

Stable Diffusion 3 Medium

Realistic images, made pretty fast. Seems like a capable and useful all-around image generator.

I like the dog it made, so if you like that dog too, maybe Stable Diffusion Medium is for you!

Stable Diffusion XL

Makes images that look fine.

I feel like this one isn’t worth spending too much time using, compared to it’s bigger-brother “Stable Diffusion 3” and the different versions of same.

I guess Stable Diffusion XL is good, but I don’t know what it’s specifically “good at”.

Stable Diffusion XL Example Image Output from Poe Ai — Stable Diffusion XL images took about 6 to 10 seconds to create

FLUX Pro

Produces highly realistic images, but takes longer to generate.

This dog was placed into a candid-photo setting, which is something that didn’t seem to be in any of the other Ai image generator’s photos of the same dog.

Even though images in FLUX Pro took about 3X longer to create, compared to Stable Diffusion 3, I think the results from FLUX Pro are worth the extra “wait” time.

FLUX Pro Example Image Output from Poe Ai — FLUX Pro takes about 17 to 20 seconds to create each image

FLUX Schnell

Super quick, photographic images.

The dog image looks like a photograph, but it also has a bit of a soft-focus/ impressionistic painting like, sometimes.

FLUX Schnell seems good, and fast.

FLUX Schnell Example Image Output from Poe Ai — FLUX Schnell took about 1 to 2 seconds to generate an image.

FLUX Dev

For super nerds to experiment with; I do not know what it’s specifically.”better” or “best” at.

I’m sure if I dive into it further, I’d find a lot of good control that I can exert over the images in FLUX Dev.

FLUX Dev Example Image Output from Poe Ai — FLUX Dev images took about 5 to 8 seconds to create

Ideogram

Great for accurately making images which feature words or text.

In the case of our dog image, I think Ideogram did a great job of making a realistic-looking dog.

If I’d wanted to add a little name-tag on his collar that said “REX”, then Ideogram would be the ideal choice for it.

Ideogram Example Image Output from Poe Ai — Ideogram images took an average of about 18 seconds to genrate

Playground V2.5

Makes colorful images that are eye-pleasing.

I’m uncertain why to use Playground V2.5, compared to others, but it made a nice-looking dog..

Playground V2.5 Example Image Output from Poe Ai — Images in Playground V2.5 took about 10 seconds to make

Poe Ai Image-Prompting Techniques:

There are a million ways to talk to an Ai image generator, but I’ve generally found the following techniques particularly useful:

Start with essentials: State your main subject before anything else (e.g., “a boxer dog”, ‘a happy person’, “a clear vase”, etc).
Add context: Use commas to add separate descriptive elements to your prompt (e.g., “Los Angeles neighborhood”, “blue cube on top of green sphere”, etc).
Specify details: Include time periods, camera angles, or specific objects in your prompt, in order to control the output image further.
Trial & improvement: Modify your prompt’s text, based on your first image’s results, and refine the output to your liking.

FLUX Pro image of a clear vase, with a green ball in the bottom, a red cone in the middle, and a blue cube on top near the mouth of the vase. — This FLUX Pro image took about 16 seconds to generate.

It’s completely possible that your first image will be perfect…

It’s also completely possible you’ll want to refine your image prompt, in order to achieve an image that matches your “ideal vision” of it.

Refining Image Prompts

Below, I’ve shared a prompt that I used in FLUX Pro, and we’ll use and modify that prompt, to make a basic ‘starter’ image:

“”A happy person walking down the street, Los Angeles neighborhood, slightly overcast day, 1930’s time period”

Here are some ways in which a small change of the text prompt can dramatically change the image output:

Different Weather Conditions:

Change “slightly overcast sky” to “night time scene”, and you’ll see how dramatically different images can result from just a few word changes.

A FLUX Pro image showing a happy person walking down the street, Los Angeles neighborhood, night time scene, 1930's time period. — This is a FLUX Pro image which took about 17 seconds to generate in Poe Ai

Changing Time Periods:

Try changing ‘1930s’ to ‘1980s’ – see how it changes the setting/ scene of your new image.

A FLUX Pro image showing a happy person walking down the street, Los Angeles neighborhood, night time scene, 1980's time period. — This image was made in FLUX Pro in about 16 seconds.

Adjusting Perspective:

Add phrases like “closeup photo” or “wide shot” or “helicopter shot” to your prompt – the new image will have a completely different perspective.

A FLUX Pro image showing a close-up photo, happy person walking down the street, Los Angeles neighborhood, night time scene, 1980's time period. — This image in FLUX Pro took about 19 seconds to generate

It’s easy to switch from a close-up to a wide shot, in an image generator; simply change a few words of text in your prompt and generate the image again.

A FLUX Pro image showing a wide-shot photo, happy person walking down the street, Los Angeles neighborhood, night time scene, 1980's time period. — FLUX Pro took about 17 seconds to generate this image.

You can create even more detailed and impressive images if you incorporate cinematic terms, or photographic terms, if you’re familiar with them.

A FLUX Pro image showing a helicopter shot looking down on a happy person walking down the street, Los Angeles neighborhood, night time scene, 1980's time period. — This image took about 18 seconds to generate in FLUX Pro within Poe Ai.

It’s not an exaggeration to say that you can change virtually anything about your generated images, provided that you have the right word to communicate your idea.

By the way, if you want some help making image prompts, you can use one of the Ai bots with Poe to get that help.

For example, all the dog images we made at the beginning of this article were generated with a prompt made by Claude 3.5 Sonnet 200K.

Using Claude 3.5 Sonnet 200K to make an image prompt — In this screenshot, you’ll see the original text + sample dog image that I gave to Claude, and the image prompt Claude made for me.

The way I made the prompt was by first feeding Claude an image of the dog I wanted to make, and I asked it to look at the image and come up with a 160-character image prompt which I can use for ai-image generators.

Then I took that image-prompt and gave it to each of the 10 image-generators in Poe Ai, and it generated the various “similar” but still “very different” dog images from earlier in this article.

This image is made of 8 different Ai Dogs — This example image from the video thumbnail is composed of 8 different samples of dog-photo; each of the photos I used is one of the examples from earlier in the article.

Text Incorporation:

When using the ‘Idiogram’ image generator, add quotation marks around text you want to appear in the image,

An Ideogram image showing a happy person walking down the street, past a "Denny's" restaurant, Los Angeles neighborhood, night time scene, 1980's time period. — This image took about 18 seconds to generate in Ideogram

Tips for Better Results with Poe Ai Image Generators:

Experiment with different models for various tasks. I love Stable Diffusion 3 and FLUX Pro for realism, and DALL-E 3 for more artistic interpretations, and Ideogram for images featuring text.
Be clear and detailed in your prompts. The more specific you are, the better results you’ll get.
Try the “turbo” versions when you need quick results, but remember they might sacrifice some detail.
Don’t expect consistent characters across multiple images – this is still a challenge for AI.

A DALL-E-3 image showing a happy person walking down the street,, Los Angeles neighborhood, night time scene, 1930's time period. — This image took about 18 seconds to generate in DALL-E-3

A Flux Pro image showing a happy person walking down the street,, Los Angeles neighborhood, night time scene, 1950's time period. — This image took about 17 seconds to generate in FLUX Pro

A Stable Diffusion 3 image showing a happy person walking down the street,, Los Angeles neighborhood, night time scene, 1950's time period. — This image took about 6 seconds to generate in Stable Diffusion 3 on Poe Ai.

A Stable Diffusion 3 Turbo image showing a happy person walking down the street,, Los Angeles neighborhood, night time scene, 1930's time period. — This image took less than 2 seconds to generate in Stable Diffusion 3 Turbo

A FLUX Schnell image showing a happy person walking down the street,, Los Angeles neighborhood, night time scene, 1950's time period. — This image took less than 2 seconds to generate in FLUX Schnell

Ethical Considerations of Image Generators

It’s important to be aware of potential copyright issues with AI-generated images.

The training data for these some of these Ai-image generators may include copyrighted images, and this has created some uncertainty around ‘ownership’ and usage rights of Ai images.

Go And Have Fun in Poe

AI image generation platforms like Po offer incredible tools for creative expression and visual communication. I encourage you to experiment, and explore the vast potential of this technology.

Make mistakes and generate less-than-perfect images – that’s all part of the learning process. With practice, you’ll develop an intuition for crafting prompts that yield amazing results.

Remember, Ai is a tool to enhance your creativity, not replace it. Your unique ideas and perspective are what will truly make your AI-generated images stand out.

Poe Ai Image Generators – ‘Stable Diffusion 3’ + ‘Flux Pro’