←back to thread

VibeVoice: A Frontier Open-Source Text-to-Speech Model

(microsoft.github.io)
448 points lastdong | 1 comments | 03 Sep 25 10:44 UTC | HN request time: 0.202s | source
Show context
Meneth ◴[03 Sep 25 14:49 UTC] No.45116481[source]▶
Open-source, eh? Where's the training data, then?
replies(1): >>45117919 #
Joel_Mckay ◴[03 Sep 25 16:47 UTC] No.45117919[source]▶
Most scraped data is often full of copyright, usage agreement, and privacy law violations.

Making it "open" would be unwise for a commercial entity. =3

replies(1): >>45120324 #
zoobab ◴[03 Sep 25 20:56 UTC] No.45120324[source]▶
Open source is being abused to not provide the actual source. Stop this.
replies(1): >>45120491 #
Joel_Mckay ◴[03 Sep 25 21:16 UTC] No.45120491[source]▶
A lot of code has multiple FOSS licenses that are not contaminating like GPL. GPL violations do occur on code, but have nothing to do with the training Data.

For example, many academic data sets are not public domain, and can't be used in a commercial context. A GPL claim on that data is often an argument of which thief showed up first.

Rule #24: A lawyers Strategic Truth is to never lie, but also avoid voluntarily disclosing information that may help opponents.

Thus, a business will never disclose they paid a fool to break laws for them... =3

replies(2): >>45121109 #>>45121115 #
1. ◴[03 Sep 25 22:35 UTC] No.45121109[source]▶