You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
What's new in 4.0.0
🖥 YTKTV Desk: Server Mode
A server console instead of the front desk. Settings → Staff Terminals → Switch to Server Mode… turns the main Desk's computer into the server: it shows rooms and staff terminals online, requests per minute, CPU, memory and database size, today's takings, and a live log you can filter and export. Staff work on the staff terminals.
It stays in Server Mode across restarts until you leave it (with the Desk password, if one is set), keeps the computer awake, can start with the computer, and backs the records up with one button.
🎮 The Desk's GPU works for every room
Process songs for the rooms: with a GPU on the Desk, it removes the vocals and writes the lyrics of the songs every room has queued. The rooms' own computers need no GPU, even a Raspberry Pi will do.
Every room's current song comes first, then every room's next one. A song sung in two rooms is done once.
When the Desk is off, busy, or cannot download a song, the room sends it the audio or does the song itself, as before.
Server Mode shows a Songs for rooms card: queue, the song being worked on, the GPU and the cache.
🎤 Lyrics that match the video
With a GPU, every song's lyrics are written on your computer by Whisper, from the singer separated from the band, so the timing matches this video. The lyrics found online are shown until then and help Whisper with the words.
Songs singing now or next are done in 30-second pieces: the lyrics appear piece by piece within seconds, while the rest of the song is still being worked on.
Without a GPU, lyrics come from online as before; Whisper writes only songs that have none.
🎚 Original and Instrumental line up
The instrumental is lined up with the original to the sample after it is separated, so switching between them no longer sounds doubled.
🔧 Also
Pull requests from forks get area labels too (GitHub restricts pull_request_target from December 2).
Fixed a CI crash in the Server Mode tests and a flaky test on Windows.
✅ Works with older versions
Rooms and Desks of any 3.x version keep working with 4.0. To have songs processed on the Desk, update both the Desk and the rooms to 4.0.
Instrumentals already in the cache line up the next time they are separated. Lyrics written before stay until the song is written again.