OneTrainer

OneTrainer is an open-source application for training diffusion models on your own machine, with dataset preparation and model previews in the same interface. It's for people adapting image or video models with their own training data. It runs on Windows, macOS and Linux under the AGPL-3.0 license.

Training options include full fine-tuning, LoRA and embeddings. Model support covers Stable Diffusion and SDXL alongside FLUX.1, Flux.2 Dev and Klein, Qwen Image, Z-Image and Ernie Image. It also works with Hunyuan Video and inpainting models, and accepts diffusers and ckpt formats. A graphical interface handles interactive work; a command-line interface supports headless training.

The dataset tools can generate captions with BLIP, BLIP2 or WD-1.4 and create masks with ClipSeg or Rembg. Masked training focuses learning on selected parts of an image. You can pair each image with multiple prompts, apply image augmentation, and train across different aspect ratios and resolutions without making every sample the same shape.

You can preview model output during training without opening another application, track progress through TensorBoard, and convert model formats through the UI. Automatic backups preserve the training state needed to resume a run. For training with EMA weights, OneTrainer can keep those weights in CPU memory to reduce VRAM use.