The trainer wall

Every bracelet on this wall is a different way to teach the painter. 20 trainers, grouped by model family. Pick by what you want to teach: a subject, a style, an edit, motion, or sound.

FLUX.1

The first generation. Mature, fast, with the largest community LoRA ecosystem.

Train Flux LoRA

subject

fal-ai/flux-lora-fast-training

The classic FLUX.1 LoRA trainer: trigger word in, character out.

7 parameters →

Train Flux LoRAs For Portraits

portrait

fal-ai/flux-lora-portrait-trainer

FLUX.1 tuned for faces: bright, detailed, prompt-faithful portraits.

9 parameters →

Turbo Flux Trainer

subject

fal-ai/turbo-flux-trainer

FLUX.1 dev LoRAs at a fraction of the usual training cost.

6 parameters →

FLUX.2

The flagship generation: higher fidelity, stronger text, edit training built in.

FLUX 2 Trainer

general

fal-ai/flux-2-trainer

The full-size FLUX.2 [dev] trainer for styles and domains.

5 parameters →

FLUX 2 Trainer Edit

edit

fal-ai/flux-2-trainer/edit

Before and after pairs on full-size FLUX.2 [dev].

5 parameters →

FLUX.2 Klein

Open-weights FLUX.2 in 4B and 9B sizes. Cheap to train, quick to iterate, used for every live example on this site.

FLUX 2 [klein] 9b Base Trainer

subject

fal-ai/flux-2-klein-9b-base-trainer

The workhorse trainer for subjects and styles on FLUX.2 Klein 9B.

5 parameters →

Flux 2 Klein 9B Base Trainer

edit

fal-ai/flux-2-klein-9b-base-trainer/edit

Teach Klein 9B a transformation with before and after image pairs.

5 parameters →

Qwen Image

Strong prompt-following image family with first-class edit LoRA support.

Qwen Image 2512 Trainer

general

fal-ai/qwen-image-2512-trainer

Train LoRAs for Qwen Image 2512, the current Qwen base.

4 parameters →

Z-Image

A lean, fast image family with turbo variants for low-latency LoRA inference. Mind the names: z-image-trainer trains Turbo; only z-image-base-trainer trains the base model.