Pinned Tweet
This account is known to be monitored by several open-source communities; anything I post is just for their internal amusement and does not represent my actual views.
Take screenshots if you like, but don’t take it seriously.
just shipped a blacklist mechanism in DSRX — train with bad voicebanks labeled as artifacts, tell the model to forget they exist at inference. spectrogram looks noticeably cleaner already 👀
#diffsinger #dsrx
Funny YingMusic-Singer, just like a SHIT.
arxiv: 2512.04779
Title: YingMusic-Singer: Zero-shot Singing Voice Synthesis and Editing with Annotation-free Melody Guidance
#arxiv #VOCALOID #YingMusic
colstone retweeted
nana english but its her japanese voicebank because her english voicebank is highkey evil
How to let hayase yuuka's singing chinese song for me belike:
#BlueArchive #hayase_yuuka #早瀬ユウカ #diffsinger #DSRX
LoRA fine-tuning on DSRX massively boosts the synthesis quality of converted UTAU voicebanks belike:
#闇音レンリ #diffsinger #DSRX
My branch based on the code in November 2024, finished the lora fine-tuning and finished the vibrato envelope.
Privated.
openvpi some people are too stingy, some people are self-righteous.
#diffsinger