[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/vt/ - Virtual Youtubers

Search:


View post   

>> No.42066849 [View]
File: 269 KB, 350x495, kronii500.png [View same] [iqdb] [saucenao] [google]
42066849

Guy who made the initial Kronii tests on /waifu/, hijacking the thread a little. I don't mean to doompost or downplay how impressive some of the results of ElevenLabs have been (especially >>42041626 ), but I'd like to remind everyone that there is not a guarantee that ElevenLabs will last forever, if even for a few months, without getting slapped by some sort of filter. There's blood in the water and "muh ethics" are already at the coattails of the site.
What everyone here is doing in regards to data collection is VERY good though. Keep your audio files in their best quality and archive fucking everything, even if you've already gotten a good Eleven model out of a chuuba. If Eleven gets fucked and we have to resort to traditional voice AI solutions (15ai/TalkNet/SoVitsSvc/etc) we will have a MASSIVE run on things by already having large amounts of cleaned data. I'd also suggest that if you're YTDLP-ing whole streams such as watchalongs for this, to also make sure to download the subtitle files using "--write-subs --write-auto-subs" to make the inevitable mind-numbingly boring transcription process easier. If you really want to get the jump on things then starting transcription work now isn't a bad idea either as long as you're not cripplingly ESL.
Other than that, I'll keep myself a lurker for now. Can't wait to see what else this general produces!

Spoken version: https://files.catbox.moe/ugxmox.mp3

Navigation
View posts[+24][+48][+96]