Repository navigation
Mildify v1.3.0
Added
- The AI DJ can write its talk with a cloud model: OpenAI, Anthropic or Google Gemini, with your own API key, which
is kept in your system's keychain. Pick the provider and model in Settings → AI DJ; the voice is still made on
your computer. - Before picking a set, the DJ can look songs up: their genres, release date and label, how popular they are, and
the artist's story, from Spotify, and from MusicBrainz if you allow it. Cloud models and Qwen3 4B do this, and so
can your own model server if you say its model supports it. - "Let the DJ name every song in a set", in Settings → AI DJ.
- Ask the DJ for its next set in your own words ("something upbeat", "more Radiohead", "the 90s") on the DJ page,
and skip the rest of a set you aren't feeling with "Skip this set".
Changed
- The DJ doesn't greet you again with every set: after the opening it carries on the show ("next up…") and doesn't
repeat what it said before. - The DJ introduces only the first song of a set, instead of reading out the whole set list.
- The Linux AppImage needs glibc 2.39 or newer (Ubuntu 24.04, Debian 13, Fedora 40 and later). On older systems,
install the .deb or .rpm instead.
Fixed
- In the Linux AppImage, sung lyrics no longer shimmy: it now comes with a newer WebKitGTK (2.52).
- With "Pick songs as it goes" on, pressing Next or Previous several times quickly no longer runs past the DJ's set
or stops the music, and Previous on a set's first song starts it again.
What's Changed
- AI DJ: cloud models, song look-ups, set requests and set skips by @hackrxd in #5
- Add claude GitHub actions 1791319328641 by @hackrxd in #6
Full Changelog: v1.2.0...v1.3.0