- Flux is an open-source AI model for generating images. It has been developed by former Stable Diffusion engineers, under a new company called Black Forest Labs.
- Compared to Midjourney, it produces highly photorealistic images and more accurate renders of human features like hands and eyes without any artifacts.
- In its current stage, it's not as feature-rich as Midjourney, but it's cheaper and easier to access.
Till now, Midjourney has been ranked as the best AI image generator, and rightly so. However, a new open-source model called Flux. 1 has been released by Black Forest Labs, and it has quickly become a sensation for generating photorealistic images. In this Flux vs Midjourney comparison, we’ll check which tool is better at generating images, depicting human features and the images look quite surreal. The realism in generated photos is pretty remarkable.
What is Flux AI Image Generator?
Flux. 1 is an open-source AI model for generating images, and it’s developed by former Stable Diffusion engineers and researchers. Under the new AI startup called Black Forest Labs, three new Flux. 1 models have been released. Flux. 1 [pro] is the state-of-the-art model, and it’s for commercial use only.
Flux. 1 [dev] and Flux. 1 [schnell] models are open-source and the AI community has already improved the models for generating photorealistic images. With a Realism LoRA adapter, the Flux. 1 [dev] model is able to generate visually realistic images and the hands are depicted without any distortion or artifacts.
So to test the Flux. 1 AI model, we have compared it with Midjourney v6.1. We have used the FLUX Realism LoRA model on fal.ai (visit) for human subjects and FLUX.1 [pro] (visit) for scenic images. Here is the comparison between Flux and Midjourney using the same prompts.
Flux vs Midjourney: Image Generations
1. A Woman Commutes at Sunrise
In the first set of images, you can notice that Flux’s generated image is more photorealistic. Despite using the --style raw parameter in Midjourney, it generated a cartoonish image.
2. An Elderly Couple in a Park
Again, Flux’s output is more realistic and believable than Midjourney’s result. Midjourney’s image looks like a painting, whereas Flux generates a lifelike depiction.
3. Image of a Chef
In this test, both images generated by Flux and Midjourney don’t pass the sniff test. Both images look AI-generated, but Midjourney’s result demonstrates better handling of reflections and lighting.
4. A Fantasy Castle
Here, I would pick Flux’s output as it depicts imagery straight out of a fantasy. Midjourney’s generated image is also pretty detailed.
5. A Futuristic City at Night
This one would be a tie, but Flux’s image is a bit clean, whereas Midjourney has generated a visually complex image.
6. A Newspaper with Legible Letters
Finally, in text rendering, both Flux and Midjourney did a great job. This is one area where image synthesis models struggle, but it’s a good sign that an open-source model is able to compete with a closed-source model like Midjourney.
Winner: Flux
Flux vs Midjourney: Pricing
Now, coming to pricing, well we know that Flux is free and open-source, but to run the model locally, you will need beefy specs with a dedicated GPU. It’s a 12 billion parameter model, after all. The best way to run Flux on local hardware is through Pinokio (website). Download the app and get started with Flux.
Apart from that, there are many service providers that allow you to generate images using the Flux model for free. However, those come with limited free credits or the image generation might take considerable time, depending on the system load.
You can head over to HuggingFace (website) to use the Flux. 1 [dev] model for free. In addition, you can check out fal.ai (website) where LoRA models are also available, but you only get $1 worth of free credits. After that, you will have to buy more credits based on your usage.
As for Midjourney, it’s expensive, and it costs $10 per month when billed monthly. But the good part is that image generation is pretty fast, and you get a slick UI to boot.
Winner: Flux
Flux vs Midjourney: Features
Since Flux is a new model in town, it doesn’t offer many AI image editing features out of the box. But there are some services like BasedLabs (website) that allow you to upscale images, and define the aspect ratio. You can also choose how many images you want to generate at once. And if you are a technical user, you can play with seed value, steps count, and de-noise settings as well.
Midjourney, on the other hand, is more feature-rich and supports many features. To begin with, you can set the aspect ratio, choose your aesthetics, and stylization. You can even use older Midjourney models, and create a natural-looking images using the Raw option.
Moreover, after an image is created, you can choose to upscale the image, zoom out and add more details, do inpainting using the lasso tool, reframe it, and much more. Basically, you can modify an image by running new prompts, which is awesome. So for inpainting and outpointing needs, Midjourney offers more tools and controls than Flux, as of now.
Winner: Midjourney
Flux vs Midjourney: Content Credentials
If you are an artist, you would want the service to support the Content Credentials standard. It helps in finding information about the origin of the image, and whether it has been modified. Sadly, neither Midjourney nor Flux support the Content Credentials standard. These services don’t add any sort of verifiable metadata in their generated AI images.
Winner: Neither
So that wraps up our comparison between Flux and Midjourney. I am very impressed with the Flux model, and it stands as a capable and free alternative to Midjourney. The open-source community can build upon it and add features like inpainting and outpainting as well. Anyway, that is all from us. If you have any questions, let us know in the comments below.
