SimpleTuner¶

SimpleTuner is a multi-modal diffusion model fine-tuning toolkit focused on simplicity and ease of understanding.

Get Started

Install SimpleTuner and train your first model in minutes

Installation
Web UI

Configure and run training through a sleek web interface

Web UI Tutorial
Local Metrics

Store scalar charts and validation comparisons with each training run

Local Training Metrics
CaptionFlow

Generate dataset captions with local GPU workers

CaptionFlow Captioning
NSFW Checks

Filter VAE cache samples with local classifier policies

NSFW Classifier Checks
REST API

Automate training workflows with the HTTP API

API Tutorial
Cloud Training

Run training on Replicate or distributed workers

Cloud Training
Multi-User

Enterprise features: SSO, quotas, RBAC, worker orchestration

Enterprise Guide
Model Guides

Step-by-step guides for Flux, SD3, SDXL, video models, and more

Model Guides
Experimental Methods

Research features such as AnyFlow, MixFlow, Explorative Modeling, NextLat, DiffusionBlocks, ConvRot-style SDNQ Hadamard quantization, segmented checkpointing, Unsloth-style checkpointing, Prompt2Effect, Self-Flow, Self-Transcendence, Flow-DPO, Internal Guidance, iREPA, LayerSync, Diff2Flow, Metal Flash Attention, Video CREPA, and Voice Cloning

AnyFlow · MixFlow · XM · NextLat · Voice Cloning · DiffusionBlocks · Self-Transcendence · iREPA · ConvRot / Hadamard SDNQ · Segmented Checkpointing · Unsloth Checkpointing · Metal Flash Attention

Features¶

Multi-modal training - Image, video, and audio generation models
Web UI & API - Train via browser or automate with REST
CaptionFlow captioning - Generate captions with local GPUs through the Web UI job queue
Worker orchestration - Distribute jobs across GPU machines
Enterprise-ready - LDAP/OIDC SSO, RBAC, quotas, audit logging
Cloud integration - Replicate, self-hosted workers
Memory optimization - DeepSpeed, FSDP2, quantization

Type	Models
Image	Flux.½, SD3, SDXL, Chroma, Auraflow, PixArt, Sana, Lumina2, HiDream, and more
Video	Wan, LTX Video, Hunyuan Video, Kandinsky 5, LongCat
Audio	ACE-Step

See Model Guides for complete documentation.

SimpleTuner is open source software.