[ home / all / search / radio / downloads ] [ leek / negi / mmd / live / c ] [ g / j / meta ] [ login / status ]

/leek/ - Vocaloid Lounge

This is the Vocaloid board, dayo

Name
Options
Comment
Emotes
Show Emotes


File Settings
Verification
File
Embed
Password (For file deletion.)



[Return] [Catalog] [Bottom]


 No.21967

How hard would it be to make a vocal synth from scratch?

 No.21968

>>21967 (me)
I mean a vocal synth software, not a voicebank

 No.21970

Ask Kanru Hua. I would guess hard.

 No.21971

Ask that QueueSevenM guy on Kissu

 No.21972

Does it even need to be specifically that? Since at bare minimum any music software and carefully arranged and edited audio files can be made into vocals. Also depends do you want the whole thing from scratch or just make a front end GUI to a backend software? Im no programmer though but did a little beginner stuff in past.

 No.21983

I’ve been interested in doing this myself too but I know Jack about code lol

Honestly trying to make a UI for an existing engine like UTAU or DiffSinger probably is good start for practice, then you can prob move into making your own.

Iirc SynthV used Moresampler as a base(?)

 No.21986

>>21983
>SynthV used Moresampler as a base
huh?

 No.21990

>>21986
Kanru made moresampler.

 No.21991

>>21990
Thought you were saying SV was built on top of the code for Moresampler

 No.21994

If I recall correctly, SynthV was re-built from scratch 5 times, and that’s why it’s V. Probably Moresampler was used for the first version. Honestly I think the auto-tuning in SynthV is more important than the audio quality though.

 No.21995

File: 1774912150915.png(31.3 KB, 600x600, 10194.png)

Don't ask them

 No.21996

>>21995
I'm still upset

 No.21998

>>21994
I thought the V just stood for Vocal not roman numeral 5. Do people in Japan even use roman numerals that often?

 No.22005

>>21998
He was in America for a while, or at least Moresampler’s hosting situation would suggest that. It was on a university website I think

 No.22013

>>21998
Fun fact, SynthV is a really clunky name in Japanese. Iirc AHS didn't like it.

 No.22015

>>22013
Too bad even before Kanru took over the name was already solidified as a brand

 No.22017

>>22013
Well what do they call it

 No.22023

>>22017
They use the English abbreviations, but it's like "shinsebui" or "esubui", which doesn't roll off the tongue like "vocalo"

 No.22711

>>21983
I think it'd be funny if we could make a Vocal synth but to be honest it feels like there's an oversaturation of synths right now: Vocaloid, SynthV, UTAU, Voisona, Cevio, ACE, VOX Factory, Maghni, Mikoto, etc.

Also no one here knows how to train a DiffSinger so I doubt anyone here would know how to code a vsynth from scratch.

 No.22712

>>22711
On second thought, Maghni probably isn't even coming out so you can take it off the loooong list of synths, but even then there's still more I haven't named.

 No.22728

>>22711
Oversaturation aside, there isn't really a niche market we could target if we were to make a synth. ACE, UTAU and Mikoto already cover community made 'loids, Maghni (if it comes out) covers the PFX ENGloids and SV, VoiSona and Vocaloid cover everything else.

 No.22737

>>22728
ACE is for AIbros

 No.22740

>>22737
and the chinese

 No.22746

>>22728
Maybe mobile users? Yeah open utau is available on android, but not iOS.

 No.22748

>>22728
Doing it for fun is also allowed

 No.22750

>>22746
Being a mobile producer sounds painful. There are no good DAWs on iOS/Android.

 No.22771

>>22748
Seconding this. Can do it just for the sake of doing it.

 No.22772

>>22771
I'm not against it but there's so many synths no one uses these days

 No.22801

I should mention there is not much point unless you do something unique like a new type of vocal synthesis method, a way to control it live or a lot of automated processes that make producing easier and without locking down creativity.

 No.22802


 No.22812

>>22801
Yeah, there's no point in making a vocal synth unless it brings something new to the table. I think an AI powered synth like SynthV with the same quality but free would be a game changer but at the same time DiffSinger exists.

 No.22818

>>22812
NEUTRINO also exists. I have mixed feelings on the machine learning "AI" synthesis method. While I like some realistic vocals if unique but I still also like the unique sound of older methods being akin to autotune, vocoder and talkbox being their own sounds.

 No.22821

>>22818
There was a thread on this board about how sample based synthesizers are dying in favor of AI ones. However i think that Mikoto Studio engine that's coming out soon is a non AI engine.

 No.22851

>>22821
Hate to break it to you but Mikoto uses AI too. It does support UTAU banks though and will likely have its own non-AI voicebanks built in.

 No.23360

>>21967
If you make a new synth make it sample based because there's an oversaturation of AI synths

 No.24130

>>21983
Idk why but this post reminded me that some company was making a vsynth but the UI was entirely vibecoded so they had to redo it from scratch lol. Forgot which engine it was.

 No.26888

UTAU already covers the sample based voicebank niche and SV covers the AI stuff.

What would you add to your synth to make it unique?.

 No.26890

>>26888
Probably XSY but I mean... UTAU V already covers that

 No.26895

>>26888
I’d make a synth that’s a clutch for people who can’t sing. Think synthesizer V and vocaflex, but one package. With multiple languages support and also different accents for the English vocals (you can make yourself sound Australian for example)

Another idea is more gimmicky, but maybe a software for AI virtual vtubers, but with focus on singing that also takes care of lipsync.

And a really bad idea, but maybe cloud-based vsynth that’s the same as any other BUT with a gacha for voicebanks. There would be limited time voicebanks too for max FOMO.

 No.26896

>>26895
You're giving bad people good ideas with that last one

 No.26903

>>26895
>Gacha vocalsynth
Anon if this ever becomes a thing it'll be your fault and I'll hate you forever

 No.26910

>>22737
ACE isn't that bad honestly, its just that their pandering to AIbros ruined their rep

 No.27139

>>26895
>Think synthesizer V and vocaflex, but one package
There's already DiffSinger and RVC, both of which are free

 No.27147

>>22023
What about Shinsego?

 No.27150

Come to think about it there aren't many western vocal synths besides Maghni, Mikoto and possibly some of the VST based synths, right?

 No.27151

>>27150
UTAU is mostly popular over here but it's made by a Japanese dude so I don't count it

 No.27158

>>27151
UTAU IS popular in Japan. It's only niche among producers.

 No.27604

Someone should grab that idea of a Western Otaku TTS and make it real. But ideally as an indie non profit project.

 No.27634

Just watched this, it gave me a new understanding of the purpose of VXB, but leaving here since they do essentially as for a call to developers, so if anyone does decide to make their own vocal synth, I would probably take away a few things from it.

Specifically, it's a presentation Yamaha gave yesterday about Omnivocal, which is essentially a rebranded vxb but exclusively for cubase. If anyone wants a specific idea for singing synth software, this is likely the route to go. The video should be timestamped, but if there are any issues, it's about 5 hours in.
https://youtu.be/JqbufmjUujU?t=18248

 No.27890

>>27604
Seconding



[Return] [Catalog] [Top][Post a Reply]

Delete Post [ ]
[ home / all / search / radio / downloads ] [ leek / negi / mmd / live / c ] [ g / j / meta ] [ login / status ]