assuming you're training on the english base model if from scratch then you need hours regardless of language i believe

i hope someone can implement this lol https://paperswithcode.com/paper/karasinger-score-free-singing-voice-synthesis

not the official implementation but portaspeech added pretrained models https://github.com/keonlee9420/PortaSpeech

some demos

this is what i replaced my third step with in my custom tacotron notebook

dont mind the inputs theyre from a model ive been training

just tested it in colab

17 seconds on cpu for "Hello There"

there

note this one on normal, small checkpoints have not been released yet

another test

aaaaand @User @User

crypto cringe spam 🥱

Sounds like the stereotypical VSCO girl laughing

deleted scam

what the fuck is this shit

also kind of the wrong channel for this type of stuff

text is "氣到咪咪貓猫" or in english "angry cat"

though yes its out of place

looks like espresso cookie

filipino orthography has near perfect correspondence (edge cases would be names and english) so it's quite fine it's just that the stresses aren't really placed in the right syllables sometimes. filipino does have a second writing system which includes diacritics to indicate where the stresses should go

someone pls make boyfriend and/or pico from fnf the musical tts voices

Wrong channel

If Uberduck Assistant goes online, go to #842056330835656745 and use /voicesuggest

I am trying to use SimSwap

but I keep getting this error

TypeError Traceback (most recent call last) in () 24 # img_a = Image.open(pic_a).convert('RGB') 25 img_a_whole = cv2.imread(pic_a) ---> 26 img_a_align_crop, _ = app.get(img_a_whole,crop_size) 27 img_a_align_crop_pil = Image.fromarray(cv2.cvtColor(img_a_align_crop[0],cv2.COLOR_BGR2RGB)) 28 img_a = transformer_Arcface(img_a_align_crop_pil) TypeError: cannot unpack non-iterable NoneType object