Ostris AI Toolkit

Ostris AI Toolkit is an all-in-one training suite for fine-tuning diffusion models — the tool a large share of the open image-model community uses to train LoRAs and full fine-tunes of FLUX, Stable Diffusion, Qwen-Image, Wan video models and more. It has gathered more than 12,400 stars on GitHub under the MIT license and was in active development in 2026, with new model support added as fast as the models appear.

Overview

Training a diffusion model normally means wrestling with scripts, configs and hardware quirks. AI Toolkit wraps the whole process in a friendly web UI (with a CLI underneath) that handles dataset preparation, captioning, training configuration, live progress monitoring and checkpoint management. It is deliberately tuned for consumer-grade hardware, so a single high-end GPU can fine-tune today’s large models, and it supports an unusually wide range: image models such as FLUX.1/2, SDXL and Qwen-Image, editing models like FLUX Kontext, video models including Wan 2.x and LTX, audio models such as ACE-Step, and even some LLMs.

Key Features

• Train LoRA and full fine-tunes for dozens of image, video and audio models
• Web UI to configure, launch, monitor and stop training jobs with a few clicks
• Optimized for consumer GPUs, with memory-saving techniques built in
• Dataset tools and example configs for fast, repeatable training runs
• Supports the newest models quickly, from FLUX.2 and Qwen-Image to Wan 2.2 and LTX
• CLI and job-file workflows for automation and headless servers
• Optional Ostris Cloud for renting GPUs when local hardware is not enough

💳 Pricing

The toolkit itself is free and open source under the MIT license. If you lack a suitable GPU, Ostris Cloud offers paid hosted training capacity operated by the project’s author.

Who Is It For

Ostris AI Toolkit is for LoRA creators, model fine-tuners, AI artists and researchers who want one reliable, frequently updated trainer for nearly every open diffusion model instead of a different script per architecture.