GitHub ↗

A wallpaper command line

Stop searching for wallpapers.
Start describing them.

A local model shapes your idea. DEAPI brings it to your screen.

Made with Flux
One guided setupPlain English promptSave or apply
irm https://flux-gen.moriatz.com/install.ps1.txt | iex

Local prompt writing is the default in v0.10.0. Download the model once with flux local install. Image generation requires a DEAPI key.

After setup, type flux a quiet observatory above the clouds at blue hour

Made for the whole screen

Documentation

Your first wallpaper, in three steps.

Install Flux, download the local prompt writer, and add your DEAPI key for images. Prompt writing runs on your computer without a prompt-provider key.

Windowsirm https://flux-gen.moriatz.com/install.ps1.txt | iex
macOScurl -fsSL https://flux-gen.moriatz.com/install.sh | sh
  1. Install Flux

    Use the installer above, open a new terminal, then download the model and runtime (about 2.50 GB plus runtime files).

    flux local install
  2. Run the guided setup

    Add your DEAPI key for image generation. Keep the default local prompt writer, then choose your image model and update preferences.

    flux setup
  3. Describe a wallpaper

    First run flux local start in a separate terminal and leave it open. Flux expands your idea locally, renders through DEAPI, then saves or applies the wallpaper.

    flux "a quiet blue-hour coast"

Windows uses NVIDIA CUDA by default; use flux local install --cpu on Windows without NVIDIA. macOS uses a native runtime. Allow at least 6 GB free disk space. First startup can take several minutes; CPU speed varies. Existing users switch to local writing when they run flux local install.

Configuration

Change keys and models whenever you want.

flux config
View settings and masked key status.
flux config key
Add, replace, or remove a key.
flux -pm
Choose the prompt model.
flux -im
Choose the DEAPI image model.
flux update
Install the latest verified release.
flux config updates
Choose automatic updates or notifications.

Models

One model writes. Another renders.

Flux local is the default prompt writer. DEAPI generates the image. Cloud prompt providers remain optional choices for existing workflows.

Local prompt writing

A prompt writer that runs on your computer.

FluxGen connects to a separately downloaded, fine-tuned Qwen text model. It expands short ideas locally, then passes the prompt to the existing image-generation workflow.

  1. Start with a short idea

    Flux selects relevant wallpaper guidance locally, without a cloud skill-selection call.

    a quiet embroidered coastline
  2. Preview the expanded prompt

    Run flux local install once, then flux local start. Preview the expanded text without an image-generation key.

    flux prompt "a quiet embroidered coastline"
  3. Generate the wallpaper

    DEAPI still creates the image and requires its API key. Local prompt writing does not make image generation local.

    flux "a quiet embroidered coastline"

The frozen model is experimental and did not pass its pilot quality comparison against the base model. It can drift from subjects or constraints; review prompts when those details matter. Read the evaluation and hardware notes. Cloud providers remain available; when local mode is selected, a server failure produces an error instead of silently sending the prompt to a cloud provider.