FLUX.1 dev
by Black Forest Labs
The model most LoRAs on the internet were trained on.
What it is
FLUX.1 dev is a "12 billion parameter flow transformer architecture" that turns text descriptions into images, and in this studio it exists for one reason above all others: it is the base that most publicly available LoRAs were trained on. If you have found a style adapter somewhere, the odds are good that this is the model it expects.
Its documentation describes "precise control over image generation through pre-trained LoRA adaptations" for applying "custom styles, brand identities, and personalized elements seamlessly", and it supports combining several LoRAs at once for more complex styling.
Its output is cleared for commercial use, like everything else in this studio.
What it is good at
Most style adapters you will find were trained against this base. That makes it the shortest path from "I found a look I like" to using it.
Multiple LoRAs combine for more complex styling, so a brand look and a character adapter can meet in one image.
Twelve billion parameters through an optimised inference pipeline, which is why it stayed the default base long after newer models arrived.
Where it struggles
- Black Forest Labs publishes no list of limitations for it, so none is invented here.
- LoRA scale matters. Its documentation suggests roughly 0.5 to 1.5 to balance how strongly the style is applied , too high and the adapter takes over the picture.
- It is the previous generation. For raw quality, multi-reference control and text rendering, FLUX.2 is ahead.
How to prompt it
Write for the adapter, not around it. When a LoRA is doing the styling, the prompt only has to describe the subject and the framing. If the look is not coming through, adjust the LoRA scale before you touch the words.
An example that shows what that means
A woman standing at a kitchen window with a mug, seen from behind at three-quarter angle, morning light, plain interior, waist-up framing.
What we offer here
Straight from the catalogue, so this list never goes stale. The price is the provider’s published rate with our markup on top, and it is on the button before you run anything.
Questions
Why use this instead of FLUX.2?
LoRAs. Most publicly available adapters were trained on this base, so if you have found a style you want to use, this is usually the model it fits. For everything else, FLUX.2 is the stronger choice.
Can I combine several LoRAs?
Yes. Its documentation supports combining multiple adapters, which is how a brand look and a character can be applied together.
What LoRA strength should I use?
Roughly 0.5 to 1.5 per its own guidance. Above that the adapter starts dominating the image rather than styling it.