November 19, 2026

Fantasi, Omarchy’s local image maker, is becoming a default app

0

Omarchy’s new Fantasi app makes and edits images with the local models omarchy-local-ai runs, plus short clips and songs, and it will ship preinstalled with the next Omarchy version.

Fantasi image maker showing a generated lighthouse with a red cabin at dusk, with image history thumbnails on the right and edit, add, copy and download actions below

Image: omacom/fantasi README (GitHub)

Omarchy is getting its own image maker. DHH merged a pull request into the quattro branch on November 18, 2026 that makes Fantasi a default app. In the PR’s words, Fantasi “makes and changes images with the local image models omarchy-local-ai runs”, so everything happens on your own machine. It’s coming in the next version of Omarchy.

The Fantasi repository only appeared on GitHub on November 18, 2026, and its README keeps the pitch simple: “Make images on your own machine with Omarchy’s local image models.” You describe an image and watch it sharpen, then change it by saying what should change, start from a picture of your own, or combine up to four.

It goes beyond still images. According to the README, picking a video model with the model button turns the same prompt into a short clip with sound, or animates the picture you’re changing. Pick a music model and Fantasi makes a song: you describe it, write the lyrics or leave them to the model, and play it back over its own waveform. Images and clips land in ~/Pictures/Generated with their prompts, and songs in ~/Music/Generated.

Fantasi uses the image model that omarchy-local-ai runs, and the README says it offers to start or set one up when none is running. It’s keyboard-friendly too: Enter makes the image, Esc stops or goes back, Ctrl+E changes it, Ctrl+N starts a new generation, Ctrl+O adds a picture, Ctrl+1, 2 and 3 switch between screen, square and portrait, Ctrl+B toggles the history, and F1 lists every shortcut.

There’s a command line as well, with fantasi generate "a calm lake at dusk" for a new image, --from to change an existing one, and --video or --music (with --lyrics) for clips and songs. And fantasi skill install, per the README, will “teach your coding agents to make images”.

As for how it gets onto your machine, the PR says new installs pull it in through omarchy-base.packages, and the package has been in omarchy-pkgs since 0.1.0. Existing machines get it from a migration, unless you removed the preinstalls, as Signed’s migration does. Removing and restoring the preinstalls now includes Fantasi too.

DHH merged a few companion changes the same day. An image model’s line in the agents panel used to offer a terminal agent that an image model can’t run; its button now opens Fantasi instead, with an “Open Fantasi” tooltip, and focuses the window if it’s already open. A follow-up PR does the same for app-launcher entries: installed image models like Z-Image Turbo and FLUX.2 Klein 4B used to get a launcher that just hung, and now it opens Fantasi on that model. A last fix makes sure the panel focuses Fantasi’s own window rather than, say, a browser tab whose title mentions it.

Fantasi is written in C++ and, according to its README, released under the MIT license.

Source: https://github.com/omacom/omarchy/pull/19353

Leave a Reply

Your email address will not be published. Required fields are marked *