[MOD NOTE: INACTIVE] MelodiaSynth - Web UTAU Frontend in development

ReticentResolve

Ancient Master of Filibustering
Supporter
Defender of Defoko
Wow, with the song playing inside of the editor, it's gonna make making usts 1000x easier.

I can't wait to see where this goes. :D
 

Tenkoh

As lazy as can be
Defender of Defoko
Thread starter
I know I should be implementing envelopes and pitch bends, which would make voicebanks other than CV work, but I don't have the time to do it right now. I still haven't figured out the other aspects of the Melodia ecosystem, such as accounts and voicebank ownership. But I guess this is going at a good pace for a hobbyist project.

Oh, I just realized that the server is actually running the resampler tasks in parallel (apparently I wouldn't be able to run those on a single thread even if I wanted to). I didn't benchmark it throughoutly, but on my PC it takes 59 seconds for UTAU to render Miku's track of Magnet, whereas Melodia takes only 16 seconds to render locally (hitting 100% CPU usage). According to my throttling tests, on a 10Mbps download connection, rendering that track on Melodia would still be 30 seconds faster than rendering with UTAU. Not to mention that it might be possible to convert the output to a compressed audio format, further reducing total waiting times.

tl;dr: Melodia can render faster than UTAU

EDIT: Does nmasao's resampler patcher still work? Haven't tried that in ages. I might benchmark with it later.
 
Last edited:

Sors

Local Guppie & UTAU Korean Advocate
Tutor
Defender of Defoko
Wow I like how this is going!
Can't wait till its done!
Good job, keep up the good work!
 
  • Like
Reactions: HIRATELIER

Kiyoteru

UtaForum power user
Supporter
Defender of Defoko
I know I should be implementing envelopes and pitch bends, which would make voicebanks other than CV work

If you need additional types of voicebanks with full permission to use in Melodia for experiments/testing, I would be happy to provide them.
 

Tenkoh

As lazy as can be
Defender of Defoko
Thread starter
Still working on envelopes. While that isn't ready, here's a new feature:
A new button has been added that allows you to play only selected notes. Quite useful for when you're just trying to get a single note right. Oh, and the BGM will also start from the first selected note, which was painful to implement so I hope it works properly.

By the way, it is being hosted again at a different domain: http://melodiasynth.ddns.net/
If someone using Safari could try it out and see if it works (specifically the playback bar), that would be great.
 

Tenkoh

As lazy as can be
Defender of Defoko
Thread starter
Just opened it in Safari and playback seems to not be working at all. Hm.
Hmm... That is not a good sign. But it did work in the past, right? Because if so, then it's something that can be fixed. Otherwise, I'll probably have to drop support for yet another browser...
 
  • Like
Reactions: Kiyoteru

WilyemAndK

✧・゚: *✧・゚:* professionally slow *:・゚✧*:・゚✧
Retired User
Defender of Defoko
Hm... Perhaps Melodia can have its own default voice bank! (also, with having Defoko/Utane Uta in there, you'd probably have to have permission from AquesTalk or something, since technically it IS their voice being used) Although, I guess that you could consider Lalita Laakkonen Melodia's main voice, huh? I don't see any other covers using her voice on YouTube so then I assume that she's a Melodia voice. Speaking of which I came up with a new name for Melodia voices: Melodies xD (geez I'm horrible at life lmao)
 

Tenkoh

As lazy as can be
Defender of Defoko
Thread starter
Hm... Perhaps Melodia can have its own default voice bank! (also, with having Defoko/Utane Uta in there, you'd probably have to have permission from AquesTalk or something, since technically it IS their voice being used) Although, I guess that you could consider Lalita Laakkonen Melodia's main voice, huh? I don't see any other covers using her voice on YouTube so then I assume that she's a Melodia voice. Speaking of which I came up with a new name for Melodia voices: Melodies xD (geez I'm horrible at life lmao)
I don't think I'll be adding a default voicebank. There are just so many types of voicebanks (CV, CVVC, Arpasing, etc) in so many languages that sticking to a single one doesn't feel right. My idea was to showcase the top rated voicebanks of each category to give the user an idea of which voicebanks are being used by others. I haven't given much thought to it though, considering how slowly this project has been progressing.
 

WilyemAndK

✧・゚: *✧・゚:* professionally slow *:・゚✧*:・゚✧
Retired User
Defender of Defoko
I don't think I'll be adding a default voicebank. There are just so many types of voicebanks (CV, CVVC, Arpasing, etc) in so many languages that sticking to a single one doesn't feel right. My idea was to showcase the top rated voicebanks of each category to give the user an idea of which voicebanks are being used by others. I haven't given much thought to it though, considering how slowly this project has been progressing.
I understand that perfectly well! Don't worry. Although perhaps there could at least be a couple of recorded voice banks that you can use just so that people have that option. Of course, we wouldn't have to worry about that until after the main parts of the editor are complete.
 

Tompole5

Momo's Minion
Hi, if this thread isn't dead yet.

I wanted to say it does work on my mobile browser, but I can only render one note at a time, and I did for a UST with Lalita (ID #1) but the note [a ら] did not come out good at all. If you need more info from Mr such as my browser, I'll be glad to help.
 

Tenkoh

As lazy as can be
Defender of Defoko
Thread starter
Hi, if this thread isn't dead yet.

I wanted to say it does work on my mobile browser, but I can only render one note at a time, and I did for a UST with Lalita (ID #1) but the note [a ら] did not come out good at all. If you need more info from Mr such as my browser, I'll be glad to help.
Thank you for your feedback. I am aware of some of the issues affecting mobile browsers, but due to the inherent difference between mobile and desktop apps, as well as technical limitations of mobile browsers, I am not planning on supporting these devices in the near future.
The voicebanks up for testing are CV only, so the note [a ら] should indeed fail to render.
 

Similar threads