Lite SuiteLite Suite
← Back to Docs

LiteImage

Standalone local AI image generation with A/B model comparison, HuggingFace Hub, face swap, and video generation

Overview

LiteImage is a standalone Electron application for local AI image generation. It runs separately from Lite Suite because it requires dedicated GPU resources. Supports Flux, SDXL, Stable Diffusion, and video models. The Electron frontend communicates with a Python backend (sd-cli.exe) for model loading and inference.

LiteImage uses the shared Lite Suite account system. The top-right account button shows your Google profile, trial status, and license tier, with sign-in and license activation available from the header and Settings.

Features

  • A/B model comparison with 3 view modes — split, overlay, and toggle
  • HuggingFace Hub browser for one-click model downloads
  • Chainable generation pipeline — compose multi-step workflows (e.g., txt2img → upscale → face fix)
  • Batch matrix — sweep across parameter combinations
  • Face swap with automatic facial landmark detection
  • Video generation from text prompts or source images
  • LoRA and ControlNet support
  • Queue management for batch generation
  • 14 MCP tools for AI agent integration

MCP Tools

LiteImage exposes 14 tools via its built-in MCP server, allowing AI agents to generate images, swap faces, create videos, and manage models programmatically.

Installation

Downloads are temporarily paused. LiteImage runs as a separate application with its own window.

Account and License

LiteImage is included in the shared Lite Suite trial and Pro license. Sign in with Google to start or restore your trial, or activate a Pro license key from Settings > License.

Usage

  1. Launch LiteImage
  2. Select a model from the dropdown (models download on first use)
  3. Enter a prompt and adjust settings (steps, CFG scale, resolution, seed)
  4. Click Generate — progress shown in the queue panel
  5. For img2img, switch to the Img2Img tab and upload a source image
  6. For video, switch to the Video tab and configure frame count

Port Map

| Port | Service | |------|---------| | 7426 | REST API (19 endpoints) | | 7450 | FaceSwap FastAPI | | 7451 | TalkingHead FastAPI | | 7452 | Training FastAPI |

Configuration

| Setting | Description | |---------|-------------| | Models Directory | Where model files are stored | | Output Directory | Where generated images are saved | | Default Model | Model loaded on startup | | VRAM Limit | Cap VRAM usage for smaller GPUs |

LoRA files go in loras/ under your models directory. ControlNet models go in controlnet/.

Requirements

  • GPU: NVIDIA with 8+ GB VRAM recommended (12+ GB for Flux models)
  • Python: 3.10+
  • Disk: 6–24 GB per model depending on architecture

LiteImage can run on CPU but generation will be significantly slower.

Troubleshooting

Backend fails to start. Verify the backend path in Settings points to sd-cli.exe. Re-running the installer can restore missing files.

Out of memory / CUDA OOM. Reduce resolution, lower batch size to 1, enable VRAM Limit, or switch to a smaller model.

Model downloads stall. Models are large files. If interrupted, delete the partial file and restart.