Repository navigation
Releases
DiffSampler v1.0.6
I found the reason the v1.0.5 was still getting flagged, there was some other jank but mainly because the DiffSinger model was embedded inside the exe file so now its stored in the model folder along side the notice
Fine, I'll give the source code, I know github has a menu to do this but doing this is easier to have each version of source code seperated:
DiffSampler-1.0.6-msvc-manifest-fix-source.zip
There we go
DiffSampler v1.0.5
Fixed low volume bug and fixed wonky overlaps
pretty simple fixes
there are STILL virus allegations so again here is the literal resampler code for you to look at yourself and compile if you're that insecure about downloading the pre-compiled version:
DiffSampler-1.0.5-fixed-source.zip
The python scripts are just artifacts from earlier versions
DiffSampler 1.0.4
False virus flags fixed, except for whatever the freak Bkav Pro is. Only Bkav Pro. WHAT IS BKAV PRO???????????
Anyway yeah
you need all that for the resampler to run, copy it all to ur resampler folder
ima try to fix the bkav pro issue too soon but im guess most yall arent using bkav pro anti virus
EDIT: THE ZIP IS FIXED
DiffSampler 1.0.3
Rendering speed massively improved. DiffSampler Lite canceled.
Third party notice:
THIRD_PARTY_NOTICE.txt
DiffSampler is a neural UTAU resampler based on DiffSinger.
dawg virus total flags this new version but i swear on my entire family line this isnt a virus im so upset there are some false flags
i have a few ideas why this is happening but ill try to patch it out in the next release
i asked claude to help fix the performance and now it got some false flags but i promise im not hacking u
i dont normally open source things but bc of the reasonable virus concern here is the code and shi
DiffSampler 1.0.2
Rendering speed improved as much as I can right now, its still painful but not NEARLY as painful as before. I changed my mind the full neural made everything worse which is why I didn't do that. I went through so many broken versions bro
Third party notices:
THIRD_PARTY_NOTICES.txt
DiffSampler is a neural resampler for UTAU based on DiffSinger.
Soon, DiffSampler Lite will come out (it’s almost done) which instead of using AI rendering uses PSOLA rendering and is much faster
DiffSampler 0.3.0
DiffSampler first public release, DiffSampler only contains a DiffSinger frontend (still very important to the core and handles most the actual rendering) currently. DiffSinger full backend will come soon.
Test with my Defoko inspired voicebank Naoko, made from the voice Naoko from the L&H TTS engine Japanese ver: