Lunera Support & User Guide

1. Overview

Lunera is a Prompt Generation app and hybrid AI image generation platform that operates in two modes:

Cloud Rendering (Gemini) – fast, no setup required

Local Rendering (ComfyUI) – full control, offline workflows, advanced features

In addition, Lunera functions as a universal prompt engineering tool, allowing prompts to be exported and used across multiple AI rendering engines.

2. Minimum System Requirements

macOS

- macOS 13+ (Ventura or newer)

- Apple Silicon (M1 or newer)

- 16 GB RAM minimum (32–64 GB recommended)

- 20 GB free storage (100+ GB recommended)

Windows

- Windows 10/11 (64-bit)

- 16 GB RAM minimum

- NVIDIA GPU recommended (RTX 3060+)

- 8 GB VRAM minimum (12–24 GB recommended)

3. Using Lunera

Cloud Mode (Gemini)

- Enter your Gemini API key

- Enter your prompt

- Optionally add reference images

- Click Render

Local Mode (ComfyUI)

- Enable Local Rendering

- Set URL: http://127.0.0.1:8188

- Load workflow JSON

- Click Render Local

4. Cross-Platform Prompt Rendering (Multi-Engine Support)

Lunera is designed to generate high-quality, structured prompts that can be used beyond its built-in rendering systems.

Supported External Engines (Examples)

- Stable Diffusion (Automatic1111, ComfyUI, InvokeAI)

- Midjourney

- DALL·E

- Leonardo AI

- RunDiffusion

- Playground AI

- Other diffusion-based or prompt-driven systems

How It Works

- Build your prompt using Lunera’s structured interface

- Apply base styles, modifiers, and composition controls

- Copy or export the generated prompt

- Paste into your target AI platform

Prompt Optimization Behavior

- Expands descriptive detail

- Enhances subject clarity and composition

- Adds cinematic, lighting, and texture cues

- Structures prompts for diffusion-based models

Important Notes by Engine

Stable Diffusion / ComfyUI: Full prompt compatibility (best results)

Midjourney: May require trimming or parameter formatting (e.g. --ar, --v)

DALL·E / Gemini: Works well, but may ignore overly complex syntax

Other Platforms: Results vary based on prompt parsing rules

Best Practices

- Keep prompts clean and readable when exporting

- Remove redundant weighting syntax if not supported

- Adjust aspect ratio and style parameters per platform

- Use Lunera as the prompt foundation, then fine-tune per engine

5. Face Swap (Local Rendering)

Modes

- Disabled

- Single Face Swap

- Dual Face Swap

Required Inputs

- Target Image – scene/body

- Face Source A – main face

- Face Source B – optional (dual mode)

6. Advanced Face Swap Controls

Recommended Settings

- Swap Model: inswapper_128.onnx

- Face Restore Model: codeformer-v0.1.0.pth

- Restore Visibility: 0.5

- CodeFormer Weight: 0.5

- Target Node Hint: target

- Source A Node Hint: source, face a

- Source B Node Hint: source b, face b

- Matching Strategy: Simple first-three order

7. Troubleshooting

Face Swap Failed

- Incorrect node hints

- Workflow mismatch

- Missing models

Output differs from ComfyUI

- Wrong model selection

- Incorrect restore settings

- Node mapping mismatch

Prompt Issues Across Platforms

- Engine may not support full prompt syntax

- Overly complex prompts may be partially ignored

- Adjust formatting for the target platform

8. Key Insight

The biggest difference between Lunera and ComfyUI is NOT the model.

It is node mapping accuracy.

For external engines, the biggest difference is NOT the prompt quality.

It is how each engine interprets that prompt.