Your own concepts, characters, and styles with Dreambooth LoRA
This playbook shows how to fine-tune the FLUX.1-dev 12B model with multi-concept Dreambooth LoRA (Low-Rank Adaptation) for custom image generation on your hardware platform. Unified memory and GPU acceleration let you keep the Diffusion Transformer, CLIP text encoder, T5 text encoder, and autoencoder resident while you train and generate.
Multi-concept Dreambooth LoRA teaches FLUX.1 new concepts, characters, and styles. Trained LoRA weights drop into existing ComfyUI workflows for prototyping and experimentation. The same path supports high-resolution training and inference at 1024px and above.
You'll have a fine-tuned FLUX.1 LoRA that generates images with your custom concepts and is ready for ComfyUI workflows on your hardware platform.
tjtoy toy and sparkgpu GPU) or your own datasetRequired:
Optional:
Use the matrix below to confirm your hardware platform, recommended default local settings, and whether multi-node applies.
| Hardware platform | OS | Memory | Recommended default local settings | Multi-node capable hardware |
|---|---|---|---|---|
| DGX Spark | DGX OS (Linux) | 128 GB Unified Memory | Docker images flux-train / flux-comfyui from playbook assets | — |
Hardware requirements
Software requirements
docker --version and nvidia-smi inside a GPU containerHF_TOKEN) with access granted on the FLUX.1-dev model card8188 for ComfyUIAll required assets can be found in this playbook repository.
assets/download.sh — Downloads FLUX.1-dev, VAE, CLIP, and T5 checkpoints into models/assets/Dockerfile.train / assets/launch_train.sh — Build and run Dreambooth LoRA trainingassets/Dockerfile.inference / assets/launch_comfyui.sh — Build and run ComfyUI inferenceassets/flux_data/ — Sample multi-concept dataset and data.toml training configassets/workflows/base_flux.json — ComfyUI workflow for base FLUX.1 inferenceassets/workflows/finetuned_flux.json — ComfyUI workflow for LoRA-conditioned inferenceassets/models/ if needed