Getting Started

From download to your first AI chat, image, and video in under 10 minutes.

System Requirements

Operating System	Windows 10 or later (64-bit)
RAM	8 GB minimum, 16 GB recommended
GPU (for chat)	Optional. CPU inference works, GPU is faster.
GPU (for images)	NVIDIA GPU with 6+ GB VRAM recommended
GPU (for video)	NVIDIA GPU with 8-12+ GB VRAM recommended
Disk Space	2 GB for the app. 5-50 GB for models depending on selection.

No GPU?

You can still use the chat features with CPU inference. It will be slower but fully functional. Image and video generation require a dedicated NVIDIA GPU.

1. Install

1 Download the installer

Download the latest release from the Releases page or from locallyuncensored.com. The file is called Locally Uncensored_x.x.x_x64-setup.exe.

2 Run the installer

Double-click the downloaded file. Windows may show a SmartScreen warning because the app is not code-signed. Click "More info", then "Run anyway". The installer completes in a few seconds.

Windows SmartScreen

This warning appears for all unsigned applications. The app is open source and the full source code is available on GitHub for review.

3 Launch the app

After installation, the app launches automatically. You will see the onboarding wizard that guides you through the initial setup.

2. Onboarding

On first launch, a setup wizard walks you through five steps. Each step takes less than a minute.

Step 1: Welcome

Click "Get Started" to begin.

Step 2: Theme

Choose Light or Dark mode. You can change this later in Settings.

Step 3: Backend

The app scans your system for local AI backends. If Ollama is installed and running, it detects it automatically. If no backend is found, click "Install Ollama" to download and install it directly from the app with a progress bar.

What is a backend?

A backend is the software that runs AI models on your machine. Ollama is the recommended option. It is free, lightweight, and supports hundreds of models. The app installs it for you in one click if needed.

Step 4: ComfyUI

If you want image and video generation, the app detects or installs ComfyUI automatically. If ComfyUI is already installed on your system, the app finds it and shows the path. If not, click "Install ComfyUI" for a one-click installation with step-by-step progress.

Step 5: Models

Choose which AI models to download. Models are split into "Uncensored" and "Mainstream" tabs. Each card shows the model name, a short description, file size, and VRAM requirement. Select the ones you want and click "Download Selected". You can always add more models later from the Model Manager.

Which model should I pick?

If you are not sure, start with Llama 3.1 8B (marked as Recommended). It is a solid all-rounder that works on most hardware. For uncensored use, try Hermes 3 8B or Qwen 2.5 7B Abliterate.

After setup, you see "You're all set!" and the app takes you to the main interface.

3. Your First Chat

1 Open a new chat

Click "+ New Chat" in the bottom left corner of the sidebar. The main area shows an empty chat with a message input at the bottom.

2 Type a message and send

Type your message in the input field and press Enter. The AI responds in real time with streaming text.

3 Switch models

Click the model name in the top header bar (e.g. "phi4") to open the model dropdown. Select a different model to switch. Each conversation remembers which model it uses.

Think Mode

Click the "Think" toggle in the message input to enable chain-of-thought reasoning. The AI shows its reasoning process before giving the final answer. Works with compatible models like QwQ, DeepSeek-R1, and Qwen 3.

Agent Mode

Toggle "Agent" in the top right to enable tool-calling capabilities. The AI can search the web, read and write files, execute code, and take screenshots. Requires a compatible model (Hermes 3, Qwen 3, Gemma 4).

4. Generate Images

1 Open the Create view

Click the paintbrush icon in the top header bar or navigate to Create in the sidebar.

2 Install an image model

If you have not installed an image model yet, go to the Model Manager (grid icon in the header), switch to the "Discover" tab, and look under "Image" models. Click "Install" on a model bundle. Recommended starting models:

SDXL -- High quality, widely supported, needs 6+ GB VRAM
Z-Image Turbo -- Fast generation (5 steps), uncensored, 8+ GB VRAM
FLUX 2 Klein -- Latest FLUX architecture, needs 8+ GB VRAM

3 Generate

Back in the Create view, select your installed model from the dropdown, type a prompt describing what you want, and click "Generate". The first generation takes longer because the model needs to load into VRAM. Subsequent generations are faster.

Image to Image

Switch to the "Image to Image" sub-tab to upload a source image and transform it with a prompt. Adjust the denoise slider to control how much the AI changes the original. Lower values (0.2-0.4) make subtle edits. Higher values (0.6-0.8) create major changes.

5. Generate Videos

1 Install a video model

Go to the Model Manager, switch to "Discover", and look under "Video" models. The VRAM tier tabs (All / Lightweight / Mid-Range / High-End) help you find models that match your GPU. Recommended:

AnimateDiff Lightning -- Lightweight, works on 4+ GB VRAM, short animations
Wan 2.1 -- High quality, needs 8+ GB VRAM
LTX 2.3 -- Good balance of speed and quality

2 Generate a video

In the Create view, switch to the "Video" tab. Select your installed model, type a prompt, and click "Generate". Video generation takes 2-10 minutes depending on the model, resolution, and your hardware.

VRAM usage

Video generation uses significantly more VRAM than image generation. If you see out-of-memory errors, try a smaller resolution, fewer frames, or switch to a lightweight model like AnimateDiff.

6. Settings

Click the gear icon in the top right corner to open Settings.

Providers

Manage your AI backends. Ollama is enabled by default. You can add cloud providers (OpenAI, Anthropic, Groq) or additional local backends (LM Studio, vLLM, KoboldCpp). Click "+ Add Provider" to see all 20+ presets.

ComfyUI (Image and Video)

View ComfyUI status, start/stop/restart it, and configure the path and port. If you use the ComfyUI Desktop App instead of the browser-based version, you may need to change the port here to match your setup.

ComfyUI settings with port configuration

Generation

Adjust temperature, top-p, top-k, and max tokens for chat responses.

Other sections

Voice -- Text-to-speech and speech-to-text settings
Memory -- Persistent memory system that remembers facts across conversations
Personas -- Create and manage AI personas with custom system prompts
Agent Permissions -- Control what tools the AI agent can use (file access, shell, web search)
Updates -- Check for new versions

7. Troubleshooting

Chat does not respond

Make sure Ollama is running. Check in Settings under Providers -- the green dot next to Ollama should be visible.
If the dot is gray or red, click "Test" to check the connection.
Try restarting Ollama: open a terminal and run ollama serve.

ComfyUI not connecting

Go to Settings and expand "ComfyUI (Image & Video)". Check the status indicator.
If it shows "Stopped", click "Start".
If you use the ComfyUI Desktop App (not the browser version), you may need to change the port. The Desktop App often uses a different port than the default 8188. Check which port your ComfyUI is running on and enter it in the Port field.
If the path is wrong, paste the correct path to your ComfyUI installation and click "Connect".

Image generation fails

Check that ComfyUI shows "Running" in Settings.
Make sure you have installed an image model from the Model Manager.
If you see out-of-memory errors, try a smaller resolution or a model that requires less VRAM.

Downloads are slow or stuck

Click the retry button next to the failed download in the Download Badge (top header bar).
Check your internet connection.
Large models (10+ GB) take time on slower connections. The progress bar shows the current speed.

Windows SmartScreen blocks the installer

Click "More info" in the SmartScreen dialog, then click "Run anyway".
This appears because the app is not code-signed. The source code is fully open and auditable on GitHub.

Antivirus flags the app

Some antivirus software flags unsigned Tauri applications as suspicious. This is a false positive.
Add the installation directory to your antivirus exceptions list.