User avatar
qualia @qualia@floofy.tech
1mo
there's an official resonite funsies-feature in the works for adding a way to do real time speech transcription and translation for all users in a session and i'm legitimately so excited to suddenly have a reasonable and not-disgusting use for a bunch of stupid old graphics cards
1
0
0
0
User avatar
qualia @qualia@floofy.tech
1mo
that goofy ass tesla k80 and its 24 gigs of VRAM may soon be put way the hell to use
1
0
0
0
User avatar
Jae resonite @jae@ap.j4.lc
1mo
@qualia don't worry too much, the goal in the end is for it to run on CPU only, and on as many devices as possible.
1
0
1
0
User avatar
qualia @qualia@floofy.tech
1mo
@jae that's a very respectable goal. i definitely wouldn't mind the ability to throw down a beefier model so i could host an Ideal Magnetohydrodynamics: The Bedtime Stories Edition session, and have a nice transcript afterwards

my (admittedly limited) experiences with CPU-only Whisper have been, uh .. poor. to put it kindly. maybe other models fare better here

but mostly i want the chance to get real goofy with hardware :D
1
0
1
0
User avatar
Jae resonite @jae@ap.j4.lc
1mo
@qualia in any case, the solution is Open-Source, so feel free to suggest changes (once it's properly out tho) or even contribute if you can :D
1
0
1
0
User avatar
qualia @qualia@floofy.tech
1mo
@jae i'll be all over it 👀
0
0
1
0