This idea popped into my mind the other day, surprisingly not while I was in the shower. I guess it was the combination of my nostalgia for Japan, people talking about daily new XMMS2 clients in #xmms2, and all the fun visual stuff (Clutter, WebKit, etc) I saw or heard of at FOSDEM’08.

I want an XMMS2 karaoke client!

It would run fullscreen on a computer or TV, playing music with XMMS2, displaying sing-along lyrics fetched from the web, on top of random or chosen videos.

The queue could be controlled by a song browser on an N810 connected to the main station over Wifi or Bluetooth, much like the neat touchscreens in real-world karaoke rooms (whether Japan is part of the real-world will be discussed in a future article).

Alright, what do we need for this project:

• An awesome music player framework: We have XMMS2!
• Vocal-less songs: either find MIDI files for songs, or (sexier) run the music through an xform that strips the vocals (buzzwords: signal processing, feature extraction, etc). The DeFX XMMS1 plugin could possibly be reused/ported, as it seems to do precisely that (and a few things more). In case of success, the next step would be to implement custom key adjustment!
• Microphone mix: mixed in using Jack or PulseAudio?
• Video playback: gstreamer, VLC, mplayer? Random videos could be fetched from your favourite *Tube website (it can’t be worse than the videos you get in Japanese karaoke — or well in the case of YouTube, perhaps it can).
• Annotated lyrics: We need lyrics with timestamp annotations to scroll the text and highlight it accordingly. Is there such a free resource online? Google seems to mostly know of “free” sites with karaoke MIDI files, in the animated-GIF-from-the-90’s meaning of “free”. The alternative would be fetching lyrics from a standard website and let users finger-scroll/trigger them by hand while the song is playing. Timestamps could then be saved from the first playback to be replayed the next time!
• Time: oops, the usual bottleneck. Anyone knows a good resource for free spare time?

The idea here would be to only rely on standard data (music, lyrics, videos), in order to not be limited to commercial karaoke CD+G or a small repertoire of horrible MIDI files. It would also be fun to hijack XMMS2 onto people’s computer by advertising an karaoke application :-)

So will I start working on such a client?

Absolutely not. Or not until the technical shortcomings (I’d love to have the vocal removal) are dealt with by Father Christmas, the yeti, or juhovh (or rafl?), whichever comes first. It might not be too big of a deal if DeFX can be reused.

We shall see!

PS: Ask your mom/gf before you experiment with the fluo paint in the living room.