* FAMILY · black-forest-labs/flux-lora
Flux LoRA Training and Inference
This family covers FLUX.1-Dev LoRA inference and a portrait-optimised LoRA trainer that produces vivid highlights and highly detailed likenesses.
* FLUX LORA VARIANTS
2 of 2* ABOUT FLUX LORA
Working with Flux LoRA
Two sides of the same workflow live here: a portrait-optimised LoRA trainer that produces custom weights with vivid highlights and fine detail, and FLUX.1-Dev LoRA inference for generating with weights you already have.
When to choose Flux LoRA
Come here when you are creating or running a custom Flux LoRA — particularly for faces. To generate with a general model instead of your own weights, use Flux 2 or Flux Dev.
Choosing between the variants
Portrait Trainer is the training step and produces LoRA weights; FLUX HF LoRA is the inference step that generates images with them. They are used in that order, not as alternatives.
* FREQUENTLY ASKED
About Flux LoRA
- Is the Portrait Trainer a generation model?
- No. It trains LoRA weights optimised for portraits; you then generate with those weights through an inference variant.
- What is the trainer optimised for?
- Portrait generation specifically, with vivid highlights and highly detailed results rather than general-purpose subjects.
- Where do I run the trained LoRA?
- Through FLUX HF LoRA here, or the LoRA-capable variants in the Flux utility family and Juggernaut Flux.