What is Logen?

Logen is a desktop studio for generating images with AI—entirely on your local machine. Describe what you want in plain English, organize your creative work into persistent project threads, and inspect every output at true 1:1 pixel resolution. There are no cloud queues, no per-image fees, and no prompts or images leaving your computer.

First Launch & Setup Wizard

When you launch Logen for the first time, the automated Setup Wizard configures your local AI stack without requiring terminal commands:

1. Detects Your GPU & VRAM: Identifies your NVIDIA graphics card and available video memory automatically.
2. Selects an Optimal Local Chat LLM (Ollama): Recommends the best local prompt-refinement model for your hardware tier:
   • Under 8 GB VRAM: qwen3.5:9b
   • 8–16 GB VRAM: qwen3.5:35b-a3b (Mixture-of-Experts with ~3B active parameters)
   • 16 GB+ VRAM: qwen3.5:27b
3. Installs ComfyUI & CUDA PyTorch: Provisions the local diffusion engine and ComfyUI-GGUF nodes inside an isolated application directory so it never conflicts with system Python installs.
4. Verifies Ollama Service: Checks whether Ollama is active and skips redundant downloads if your chosen LLM is already present.
5. Downloads the Default Starter Image Model: Pulls FLUX.1-schnell Q4_K (Apache-2.0) along with its required CLIP and VAE encoders.

The Studio Interface & Live System Status

Logen organizes your creative workflow into five primary views accessible from the left Sidebar:

Live System Status Indicator (Bottom of Sidebar)

Click the status LED at the bottom of the left Sidebar at any time to open the System Status Popover, which reports live health for:

• Local Engine: Python sidecar API readiness.
• Ollama LLM: Active language model or a one-click Start Ollama button if the service is stopped.
• ComfyUI Backend: Shows Running or Standby (auto-start).
• GPU & VRAM: Displays your detected NVIDIA GPU model and total video memory.

Chat & 1-to-1 Project Threads

Every chat thread in Logen is bound 1-to-1 to a persistent Project on disk:

• + New Chat: Click + New Chat in the top-right of the Chat header to start a fresh project brief with a clean AI context without losing earlier conversations.
• Project / Thread Switcher: Use the project dropdown in the top-left of the Chat header to switch between projects, inspect message and render counts, or delete old threads.
• Auto-Naming & Inline Rename: New threads automatically derive a project title from your first prompt. Click the Pencil (Rename) icon beside the project title to rename it—this automatically updates the project folder name across all associated images in your Gallery.
• Project Gallery Shortcut: Click Project Gallery in the Chat header to jump straight to that project’s filtered view in the Gallery.

Prompting & Generation Controls

• Multi-Line Prompt Composer: Type or paste complex multi-line prompts freely. Press Enter to send or Shift + Enter to insert a new line.
• Aspect Ratio Selector: Choose 1:1 Square (1024×1024), 4:3 Landscape (1024×768), 3:4 Portrait (768×1024), or 16:9 Wide (1024×576) directly in the Chat header.
• Live Stage Telemetry & Stop Button: During generation, Logen displays real-time progress (Connecting to Ollama…, Planning image generation…, Loading diffusion model into GPU…, Sampling image (3/4)…, Decoding final image…). Click Stop beside the composer at any time to cancel a run.

Self-Healing Ollama & Error Recovery

If the local Ollama service is asleep when you submit a prompt, Logen automatically spawns ollama serve silently in the background and retries your generation. When manual intervention is needed, Logen replaces generic errors with an actionable diagnostic banner:

• Identifies the exact cause: Ollama offline, Selected Ollama model missing, ComfyUI unavailable, or GPU VRAM full (OOM).
• Provides one-click recovery buttons (Start Ollama, Download Model, Settings, Retry) directly inside the banner.
• Automatically restores your exact prompt text into the composer so you never lose work on a failed run.

Project Gallery & Full-Screen 1:1 Lightbox

Click Gallery in the Sidebar to browse all locally generated images stored in Documents/Logen/images/:

• Projects Filter Bar: View All Projects or click any individual Project Folder card (displaying its cover thumbnail, title, and total render count).
• Round-Trip to Chat: While viewing a project folder, click Open Project Chat Thread to jump straight back into that project’s conversation and continue iterating.
• Instant Search: Filter renders in real time by prompt keywords, project title, or workflow/model name.

Using the Full-Screen 1:1 Lightbox

Click any image in the Gallery, Image Panel, Chat bubble, Workflow Studio, or Batch Panel to open the full-screen darkroom Lightbox:

• 1:1 Full Resolution vs. Fit to Screen: Click 1:1 Full Resolution (or double-click the image) to inspect pixel grain at native 100%–400% zoom.
• Keyboard Navigation: Press ← and → arrow keys to step rapidly through all renders in the active project, and press Esc to close.
• Bottom Action Bar: One-click actions for Copy Prompt, Open Project Chat, Copy Image (to system clipboard), Save As…, Show in Folder (reveals in Explorer/File Manager), and Delete.

2-Column Workflow Studio & Batch Mode

Click Workflows in the Sidebar to open the two-column Workflow Studio:

• Bundled Production Pipelines: Includes flux_schnell_txt2img, flux_schnell_img2img, sd15_txt2img, sd15_img2img, flux2_klein_txt2img, z_image_turbo_txt2img, sd35_txt2img, and esrgan_upscale (4x RealESRGAN), plus custom ComfyUI API JSON import in Pro.
• Humanized Controls & Inline Presets: Tune Positive/Negative Prompts, Sampling Steps, CFG Guidance, Sampler, Scheduler, Width, Height, and Seed on the left while watching the sticky live preview on the right. Click Save Preset to store reusable parameter configurations.
• Image-to-Image Input: Drag and drop any PNG, JPEG, or WebP source image into the upload zone for img2img and upscaling workflows.
• Batch Generation (Logen Pro): Switch to Batch to queue 1–16 variations with Random or Increment seeds. All batch outputs save automatically to your Gallery.

Hardware-Aware Model Manager

Click Models in the Sidebar to browse and download diffusion checkpoints. Logen displays your detected GPU and VRAM in the header and tags compatible models with a green Fits your GPU badge:

Note: Logen Pro licenses Logen application features, not third-party model weights. Always review the specific model’s upstream license before commercial use. FLUX.1-dev and Qwen-Image-2.1 are disabled in this commercial build.

Where Your Files & Models Are Stored

Logen separates your creative projects and images from the underlying runtime engines so your work is easy to back up:

1. User Assets, Projects, Presets & Outputs (Documents/Logen/)

Windows: %USERPROFILE%\Documents\Logen\ · Linux: ~/Documents/Logen/

2. Runtime Engines & Model Checkpoints (AppData)

Windows: %LOCALAPPDATA%\Logen\ · Linux: ~/.local/share/Logen/