Making the move to Flux Pro.
I went with Flux Realism because my use case is mainly human images for use in blogs, social media posts and landing pages. A typical marketer's scenario.
I also set the default image size to 16:9 because over in Midjourney I spent a lot of time copying that --ar aspect ratio line to resize my images to rectangle (hero) size, rather than the default square (1:1).
The other models are:
Flux-Schnell: This model prioritises speed, making it the fastest among the Flux series. It is designed for rapid image generation at the expense of some advanced capabilities, ideal for quick prototyping and personal projects
Flux-Dev: An open-source, community-focused model. While it’s not as fast as Flux-Schnell, it offers high-quality output for non-commercial purposes. It’s a good option for developers and artists who want to build or experiment
Flux-Pro: The most advanced and feature-rich model in the Flux series. It’s designed for commercial use and excels in detailed image generation, complex prompt following, and high-quality outputs. This model is accessible through API for enterprises
Flux-Realism: Specialised for generating highly realistic images. This model focuses on producing outputs with photorealistic detail, making it useful for industries like media, advertising, and film
It's worth noting the pricing of the different models as they use up different amounts of credits per image.
If you're trying to conserve credits, then probably best you hang around in Schnell. But with subscriptions starting at $5 for 100 credits, I'm willing to spend at the moment and get the best output for my needs.
After I got set up, I explored.
Image to Image is the first option in the dashboard. This is where you can upload an image, and give it a prompt to add or remove a character from the image, or you can use the prompt to change the style of the image. Use this option if you have a good base image.
I took one of a busy office scene and at first I prompted it to add Santa Claus to the scene.
The first attempt looked like something Dall-E might do:
I tried again, this time giving it the prompt ''Busy office scene, santa claus is in the room, gritty black and white photography, in the style of 1950s film noir''
This time it created something a little better. Definitely something to work with. I'll keep this Image to Image prompt in mind as it could be easier to edit an image, rather than create from new every time.
I don't think I'll bother with this.
Point to note: this route prefers square (1:1) images to start off with!
When you have the correct sized image uploaded, it reverse engineers the image back into a prompt. Handy if you have a pic you like and want something similar.
The thing I didn't like was that it cost 3 credits to do this! I'll be taking my images over to ChatGPT and getting it to give me the prompt - for free!
When I took the prompt Flux Pro created for me and ran it, I got an image entirely different in vibe to the one that started out!
Something I found a bit concerning is that our website wouldn't upload the pic created, said it was a corrupt file. I suspect this is a Webflow issue because I was able to upload the pics to another Wordpress site.
So Wordpress 1, Webflow 0!
I turned to my trusty coaching assistant, Perplexity, to get an explanation on Flux Lora. It gave me this.
I think this is one for the developers, I'll ignore this feature, for now.
This area looks to me like where Flux take on Canva. If you're familiar with Canva, you'll know 'elements'. Flux Designer is like elements, except instead of browsing and selecting, you can dictate and create.
One of the 'Midjourney' killer functions of Flux Pro is that it can render text without adding Greek or Russian letters, and with correct spelling. This is definitely something that many marketers will find most beneficial as the first question I get asked when demo'ing Midjourney in our AI for Marketing course is 'how can I get my client logo and text on it'?
Now with Flux Pro, you can - in theory!
I had a go at getting it to create a button, I gave it our colour reference but it went too far pink. I also think the font is a bit basic.
I could probably get it to do what I want by prompting it better, but for now I'll be sticking with Canva for elements.
What I will miss from Midjourney is the initial 4 images that it creates. I love the nuances and variations. I like that I get to choose.
For now, I will continue on my learning journey with Flux Pro. BRB with any more insights!
© Copyright 2024 AI Institute. All Rights Reserved. Terms & Conditions | Privacy Policy | Ai Policy
