Run Uncensored AI Chat + Image Gen Locally in 5 Minutes (No Docker, No Cloud)

#beginners #tutorial #ai #opensource

i keep seeing the same question in every AI subreddit: "how do i run AI locally without sending my data to openai?"

the answer used to be complicated. install ollama, configure it, then install comfyui separately, figure out the python dependencies, download models manually, pray nothing conflicts.

now its actually 5 minutes. here's how.

What You're Getting

By the end of this you'll have:

AI chat with uncensored models (no content filters)
Image generation (stable diffusion, flux, whatever you want)
Video generation
All in one UI, all running on YOUR machine

No cloud. No API keys. No subscriptions. No "we updated our privacy policy" emails.

Prerequisites

You need two things installed:

1. Ollama (for chat)

# Mac/Linux
curl -fsSL https://ollama.ai/install.sh | sh

# Windows - download from ollama.ai

Pull a model:

ollama pull llama3.1

2. ComfyUI (for image/video gen)

Follow their install guide. Its basically clone the repo and run python main.py. You'll also want at least one image model - grab Stable Diffusion or Flux from civitai or huggingface.

Install Locally Uncensored

git clone https://github.com/PurpleDoubleD/locally-uncensored
cd locally-uncensored
npm install
npm run dev

Open http://localhost:5173 and thats it. You should see the chat interface.

First Things to Try

Chat: Pick a persona from the sidebar (theres 25+ built in) and start talking. Try the "Uncensored Assistant" for zero content filtering. Or "Roast Master" if you want to feel something.

Image Gen: Switch to the Create tab, type a prompt, pick your model, hit generate. The app talks to ComfyUI in the background so you dont need to touch comfyui's node editor.

Model Manager: Go to the Models tab to see whats installed, pull new models, or delete ones you dont use anymore. No more running ollama list in a terminal.

"But My GPU Sucks"

Valid concern. Here's what actually runs on different hardware:

No GPU (CPU only): Chat works fine with smaller models (phi3, gemma2). Image gen will be slow but functional
6-8GB VRAM: Most 7B chat models + SD 1.5 for images. Solid experience
12GB+ VRAM: Basically everything. 13B models, SDXL, Flux, video gen

The app doesn't care what hardware you have. It just talks to Ollama and ComfyUI which handle the actual inference. If those work on your machine, the app works.

Why Not Just Use Open WebUI / LM Studio / ChatGPT?

	Locally Uncensored	Open WebUI	LM Studio	ChatGPT
Chat	yes	yes	yes	yes
Image Gen	yes	no	no	yes (cloud)
Video Gen	yes	no	no	yes (cloud)
100% Local	yes	yes	yes	no
Uncensored	yes	no	no	no
No Docker	yes	no	yes	n/a
Open Source	yes	yes	no	no

the TLDR is: nothing else combines chat + image + video gen locally in one app. if you only need chat, Open WebUI or LM Studio are great. but if you want the full creative suite running on your machine, this is it.