🎨 Hanzo Painter

AI-powered content-aware inpainting for videos and images

Quick Start Guide

Get Hanzo Studio running in minutes.

Prerequisites

  • Python 3.8+
  • CUDA-compatible GPU (8GB+ VRAM recommended) OR Apple Silicon Mac (M1/M2/M3/M4 with MLX)
  • Git, wget

One-Command Setup

make all

This will:

  1. Install Hanzo Studio
  2. Install all Hanzo custom nodes
  3. Download SAM2 models
  4. Install the inpainting workflow

Step-by-Step Setup

1. Install Hanzo Studio and Nodes

make setup              # Install Hanzo Studio and all custom nodes

2. Download Models

make download-models    # Download SAM2 models

You also need to manually download these models:

  1. realisticVisionV51_v51VAE.safetensors
    • Download and place in Hanzo Studio/models/checkpoints/
  2. pcm_sd15_smallcfg_2step_converted.safetensors
    • Download and place in Hanzo Studio/models/diffusers/

3. Install Workflow

make install-workflow   # Copy workflow to Hanzo Studio

4. Start Server

make run               # Standard mode
# OR
make run-mlx           # MLX acceleration (Apple Silicon)
# OR
make run-lowvram       # Low VRAM mode
# OR
make run-cpu           # CPU-only mode

5. Open Browser

Navigate to http://localhost:8188

Your First Inpainting

1. Prepare Your Video

Place your video file in Hanzo Studio/input/

2. Load Workflow

  1. Open http://localhost:8188
  2. Load workflow: workflows/inpainting-workflow.json

3. Configure

Update the VHS_LoadVideo node with your video filename:

"video": "your-video.mp4"

4. Adjust Parameters (Optional)

Parameter Default Description
num_inference_steps 15 Denoising steps
guidance_scale 10 Removal strength
video_length 50 Processing chunk size
frame_load_cap 300 Max frames to load
frame_rate 8 Output FPS

5. Run

Click “Queue Prompt” in the UI

6. Get Results

Output video appears in Hanzo Studio/output/

Optional Enhancements

Install SAM2 for Advanced Segmentation

make install-sam2

Install MLX for Apple Silicon Acceleration

make install-mlx

MLX provides:

  • 70% faster model loading
  • 35% faster inference
  • 30% lower memory usage

Next Steps

Makefile Commands Reference

make help               # Show all commands
make setup              # Complete setup
make all                # Setup + workflow + models
make run                # Start Hanzo Studio server
make update             # Update Hanzo Studio and nodes
make test               # Test installation
make clean              # Remove caches
make info               # Show installation info
make models-info        # Show downloaded models

← Back to Home Installation Guide →