CNNCT Image

YOUR COMPUTER. YOUR CREATIVE SPACE.

Make the image
you have in mind.

Create from a prompt. Edit your own images. CNNCT Image gives local AI a workspace that feels like an app.

Get CNNCT Image

For Apple Silicon and Windows with NVIDIA.
No account. No per-image credits.

CNNCT Image Generated locally
A red ceramic teapot on a wooden table beside a window
01 / CREATEA red ceramic teapot, in morning light.
The same ceramic teapot edited to cobalt blue
02 / EDIT“Change the teapot to cobalt blue.”

Real Qwen Image 2.1 outputs from an M1 Max. 512 × 512, 12 steps.

A LOCAL IMAGE WORKSPACE

From one idea
to a whole folder.

01

Write it. Make it.

Describe a new image, or choose up to three references and explain what should change. Browse the results in your library and save a copy wherever you need it.

02

Leave a queue running.

Import a list of prompts or apply one instruction to a folder of images. Jobs run one at a time. Pending work is saved when you close the app.

03

Use the same engine in your scripts.

An authenticated localhost API handles generation, edits and queue controls. Use the desktop app when you want to see the work, and scripts when you want to repeat it.

LOCAL BY DESIGN

Your prompts and images
stay in your workspace.

Model setup downloads files from GitHub and Hugging Face. After setup, generation uses the models on your computer. There are no hosted inference calls, analytics or automatic updates in the app.

For an enforced offline session, disconnect your internet connection. Keep your data folder outside cloud-sync services. The app does not install a firewall or encrypt your files.

CNNCT IMAGE 0.1.0

Give your ideas
some space.

Download the app, install the model once,
then start making images.

Before you download

Space & memory
About 11 GB of model downloads on first launch. Allow 15 GB free disk space, plus room for your images. 32 GB system memory is recommended.
Mac
Metal acceleration on Apple Silicon. Generation and editing tested on an M1 Max with 32 GB RAM. Intel Macs are not supported.
Windows
Designed for NVIDIA cards such as the RTX 3080, with an 8 GiB GPU budget. Requires a current NVIDIA driver and the Microsoft Visual C++ x64 runtime. CUDA is bundled; no toolkit installation needed. Physical RTX 3080 testing is still pending.
A first release
Powered by quantised Qwen Image 2.1 through stable-diffusion.cpp. Larger images take more time and memory. Keep your computer awake for long batches.