Exploring FLUX.1: A Quick Guide with Comparisons and Walkthroughs
Written on
Getting Started — Let's engage in an interactive activity that we refer to as...
Game Instructions: Below is a grid featuring nine images of miniature “tiny lions,” all created using the following prompt:
A close-up shot of a micro tiny realistic adult lion perched on a human finger, with intricate and extreme details, shot with a Sony A7R IV, macro lens
Out of these, five images were produced using the FLUX.1 model, while the other four were generated by different models.
Can you identify the five images created with FLUX.1?
For a clearer view, here are enlarged versions of the nine images.
Tiny Lion — Image 1
Tiny Lion — Image 2
Tiny Lion — Image 3
Tiny Lion — Image 4
Tiny Lion — Image 5
Tiny Lion — Image 6
Tiny Lion — Image 7
Tiny Lion — Image 8
Tiny Lion — Image 9
Were you successful in identifying the five images created with FLUX.1? Which clues led you to your conclusions? What traits assisted you in discerning the FLUX.1 images?
Answers to the FLUX.1 or NOT? Game.
- NOT FLUX.1
- FLUX.1
- FLUX.1
- NOT FLUX.1
- NOT FLUX.1
- FLUX.1
- NOT FLUX.1
- FLUX.1
- FLUX.1
How did you fare? Were you able to find all five FLUX.1 images?
> Note: Honestly, this game was quite challenging. I generated the images across nine different platforms and labeled each image file on my computer with its origin. If I hadn’t labeled them, I would have struggled to determine which were from FLUX.1 and which were not. I found many images so similar that distinguishing them was difficult. While I did notice some that were more appealing than others, identifying them as FLUX.1 or not was a challenge.
Now, let's take another look. The image below displays only the tiny lions generated by FLUX.1, along with the sites where they were created.
Next, we will briefly explore each platform, but first, a bit of context.
What is FLUX.1?
FLUX.1 is an AI image generation tool developed by Black Forest Labs (https://blackforestlabs.ai/), created by the four co-founders of the original Stable Diffusion. It is available as open-source software, allowing users to download and install it for free on personal computers. Developers can inspect and modify the source code to improve the software or create their own variations.
FLUX.1 has three distinct versions (or models):
- FLUX.1-pro
- FLUX.2-dev
- FLUX.3-schnell
All three models are scaled to 12 billion parameters (12B), which is a substantial amount of data used for training the models, with the pro version being the most expensive yet offering the highest creative potential.
The pro version is the largest and most robust of the three. The dev version is nearly as powerful but smaller and quicker in generating images. The schnell version (meaning "fast" in German) is the most efficient, producing images rapidly, though they may be of slightly lower quality. The schnell version is typically free, but using the pro and dev versions may require account setup and could incur a small fee per generation (typically in thousandths of a cent, e.g., $0.00025 per image).
FLUX.1 is commonly deployed in two ways: either integrated into existing products or provided as a standalone service by an AI provider. We will see both types as we go through the walkthroughs.
Walk-through #1 — FLUX.1 as part of the FREEPIK product
Freepik offers a limited free tier allowing a specific number of AI generations and downloads daily. For more details, visit their website at https://www.freepik.com/.
- Begin with your Prompt Text.
- Select the Model (three FLUX options).
- Set the Size (aspect ratio).
- Choose a Style (optional — I didn't select one in this instance).
- Click Create.
Here’s the outcome (it generated four images).
From here, you can click to Upscale and make further edits in the Pikaso (FreePik) Editor.
Walk-through #2 — FLUX.1 as part of the DZINE product
The pricing page at https://www.dzine.ai/pricing lists a free account option that grants 30 credits daily.
- Click New Project.
- Enter a Project Name.
- Choose an Aspect Ratio (or input custom dimensions).
- Click the Apply button.
- Select the Txt2Img icon, and then choose FLUX.1 as the style.
FLUX.1 will appear in the left sidebar.
- Enter your Prompt Text.
- Finally, click the Generate button.
Two thumbnail images will appear in the right sidebar. Hover over each thumbnail to see a larger version.
- Double-click the selected image to load it into the Canvas for further editing.
Walk-through #3 — FLUX.1 as part of the NIGHTCAFE STUDIO product
You can access NightCafe Studio at https://creator.nightcafe.studio/ for free, although the free account is limited to a set number of credits.
- Click the Create button on the homepage, which will direct you to the Create Screen.
- On this screen, click the Model button. A list of models will appear on the right side.
- Select the Flux model (the Flux Schnell model is also available in the second row). Flux will appear in the left menu.
- Input your Prompt Text.
- Click the Preset Style button. A list of styles will appear on the right. Adding a style is optional; for this example, I will choose None as the style.
- Click the More Settings button to specify the number of images to generate (this model allows only one at a time) and to Change the Aspect Ratio.
- Click the Create button to generate the image.
Here’s the result:
You can see additional options at the bottom of the image to upscale or continue editing.
Walk-through #4 — as an AI model on HUGGINGFACE
The Hugging Face website hosts numerous AI and Machine Learning Models. You can find the three FLUX.1 models there. The Dev and Schnell models are generally free to use, while the Pro model incurs charges per generation.
Visit https://huggingface.co/spaces/black-forest-labs/FLUX.1-schnell. I managed to run this without logging in (anonymously).
- Enter your Prompt Text.
- Click Advanced Settings (to expand the tab and set a seed value, adjust the aspect ratio by specifying width and height, and modify the number of inference steps. All these settings are optional).
- Click Run to generate the image.
A progress bar will appear, and once completed, your image will display. Here’s the result:
- Click the small icon to download your image.
Walk-through #5 — as an AI model on REPLICATE
Replicate is another platform that features a variety of AI and Machine Learning Models. You can also access the three FLUX.1 models here. The Schnell model is typically free to use, while the Dev and Pro models incur charges per generation.
- Enter your Prompt Text.
- Choose your Aspect Ratio (the dropdown will display options). You can set other optional configurations.
- Click and set the Output Format for your image.
- Click Run to generate your image.
Here’s the output.
- Click the Download button to save the image you created.
> Note: The Replicate platform stated that you could utilize its featured models for free, and FLUX-schnell was listed as a featured model — however, the site required me to log in with my Github account before I could generate anything — proceed with caution!
Here are the four tiny lions generated without using FLUX.1.
- Midjourney AI — https://www.midjourney.com
- Ideogram AI — https://ideogram.ai
- Recraft AI — https://www.recraft.ai
- Leonardo AI — https://leonardo.ai
Some of these platforms are now only accessible to paid subscribers, while others may offer limited versions for free.
That concludes my article.
I hope you found this information useful and can apply some of the insights shared here. All my articles are accessible via my Medium profile. Please clap (up to 50 times) if you enjoyed reading. You can also follow me and join my email list to receive updates on all my articles!
Best, David
This article is published on Generative AI. Connect with us on LinkedIn and follow Zeniteq to stay informed about the latest AI developments.
Subscribe to our newsletter and YouTube channel to keep updated with the latest news in generative AI. Together, let's shape the future of AI!