SimpleTuner is an open-source fine-tuning toolkit for diffusion models, covering image, video and audio generation. It provides a single, configuration-driven training environment for models such as Stable Diffusion, SDXL, Flux and popular video models, supporting everything from quick LoRA adapters to full model fine-tuning across multiple GPUs. The project has roughly 3,000 stars on GitHub and is updated very actively, tracking new model releases closely.
Fine-tuning diffusion models usually means juggling a different trainer for every architecture, each with its own quirks and dataset format. SimpleTuner standardizes that workflow: datasets are described in a unified configuration, captions and metadata are handled consistently, and training options โ learning-rate schedules, memory-saving techniques, distributed setups โ are exposed the same way regardless of the target model. It emphasizes reproducible, scriptable training for people who fine-tune often, and it adds conveniences such as automatic captioning helpers, validation previews during training, and careful VRAM management so large models can be trained on consumer hardware. Extensive documentation and example configurations for popular models lower the learning curve for newcomers, while experienced trainers can override nearly every aspect of the process.
SimpleTuner is free and open source under the AGPL-3.0 license. There is no paid version or hosted training service โ you run it on your own GPUs or rented compute.
SimpleTuner is for AI art practitioners, model trainers and researchers who fine-tune diffusion models regularly and want one consistent, scriptable toolkit instead of a separate trainer per architecture.