kohya

Turn a character, style, or concept into a ready-to-train LoRA dataset - consistent multi-angle shots, smart captions (prose / Danbooru / e621 / dedicated taggers), and trainer-ready configs for Flux, SDXL, Krea, Pony & more. Local or cloud, per stage. No training launched, no telemetry.

Bring a folder of images, leave with a trained LoRA. Agent skills (SKILL.md) for quality-gated LoRA training: dataset doctor + safe fixer + one-confirmation training orchestration for SD-Trainer / lora-scripts-next. Anima-first, SD1.5/SDXL/Flux.

Automated Stable Diffusion LoRA pipeline with dataset extraction, training, generation, and evaluation.

LoRA training and sampler ablation study for Stable Diffusion — IEEE 2024

Complete sovereign LoRA pipeline. Pinterest to trained SDXL LoRA in 2 hours on a single GPU. Zero cloud APIs. Includes RAM RAID image extractor, LLaVA auto-captioning, and battle-tested Kohya training config.

Automate LoRA model training with a guided pipeline for curation, tagging, and validation compatible with Claude Code and other agentic workflows.

Local LLM + ComfyUI + Open WebUI on a single NVIDIA GPU. llama.cpp proxy with auto GPU mode swapping (LLM / image-gen / LoRA training). Benchmarking + model-preset switching.