Hi
I have a number of tunes In BIAB 14 where lyrics are carefully synced to the midi melody. Have tried the "Vocal Synth" with several of these but only achieved a sound that is somewhere between a cat and a muted trumpet. with no words. BIAB sent me to the synth's web site at www.sinsy.jp but couldn't figure out what to do next. I did hear some impressive demos though, would love to get it working. It (she) sings better than I do.
Help here would be greatly appreciated....Hank

If those songs don't have the Note Based Lyrics but instead have the older lyrics that appear above the notation in Notation View, I don't think they will work with this feature.

You would have to first have a Melody entered on the Melody track that follows the lyrics, then enter the lyrics as Note Based in order to do it.


--Mac

Perhaps try to find and work with a real singer? Those synth things are generally pretty lousy.

Hi Mac
When You mention note based lyrics, I presume you mean each word or syllable is placed according to the correct note. The lyrics I am using were entered this way.
Thanks for your reply....Hank

I hadn't tried it out before, but I get exactly the same results by using Vocal Synth - Auto Mode. The syllables all sound like "kih".

However, Vocal Synth - Manual Mode works OK. After creating the file, you can manually load it on the Sinsy site, then choose wav and save that for loading into BiaB.

https://www.youtube.com/watch?v=B7BJIsMZkYo&feature=youtu.be

Also, there are two different vocalists listed, in my testing I found that the Chinese one did a much better job at the English lyrics than the Japanese one, YMMV.

I thank you all for suggestions
Jazzman, I have viewed the video but It's not working correctly for me. When I go to sinsy.jp and click where you choose file, a window headed "Choose file to upload" opens and every folder in my computer is listed there, so I have no idea what to do next....Hank

You must locate the bb songfile you want her to sing, it must be .mgu type, and select it from there.


--Mac

Wow! I listened to that and the generated vocals were...awful! What in the world would anyone use this for? It wasn't just that it sounded artificial...many of the words/syllables were not even remotely correct. And, because PG used that example as their demo I am assuming this is as good as the current technology gets?

Seriously, what would anyone use this for?

Hi All,

Take a listen to this and see what can be achieved (pity PGMusic can't replicate)

Vocal Synth

George

Hi Mac, as i was Curious as a Cat, i tried all that, and guess what, LOL: ... it indeed sounded like one of our furred friends ... But if you can get something more out of it, you got me as i am curious again ! - F

The thing is in its infancy, who knows at this point in time if it will develop into something usable in production, or perhaps into something along the lines of a Special Effect sort of thing someday, as in the overuse/abuse of Autotune to get the Cher Effect was.

But consider that there are some folks who cannot speak or sing anymore, like our own good forum member Don, who just might be able to get enjoyment out of their songwriting and demos from the use of this feature.

So many on this forum only see Band in a Box as whatever it is that they themselves want to use it for, there's plenty of room for all the other stuff as well, if you don't like a particular feature, style, addon or command in the program, just don't use it. There are likely others who view that same thing as their "must have".


--Mac

with a software version of TC helicon available here in BiaB why even try that. Just sign a very simple none vibrato basic vocal track, and generate a harmony and have it fill out separate tracks, and then mix and texture to taste.

For those who question why anyone would want this vocal synth. When I'm playing one of my very rare gigs, I'm backing my vocalist along with my RB made backing tracks. At home I practice with the same backing tracks, substituting piano or guitar etc. where the vocalist would normally be. I just thought it would be nice to have a vocal at those times....Hank

PS That youtube demo is fantastic. thanks.

Yeah, I certainly appreciate that sort of use and I have heard computer generated voices that were reasonably accurate but still artificial sounding...these would be great for someone who cannot speak or sing. But that demo was terrible and many of the words were not even remotely correct. Why would anyone want something so bad? You could hit a garbage can lid with a stick and get closer on many of the words!

Thanks for the video...that one was a lot better. I wonder if they did a lot of work on that one though? Maybe the automatic stuff is just still a long ways off from getting the words right.

During the beta testing I ran "Amazing Grace" thru Sinsy as a test.

First time through, using Note Based Lyrics "as written" got some pretty bizarre pronunciations.

"Ah - Mozz - in - Grauss" kind if thing,.

Rewriting the Lyrics such that the spelling forced Sinsy to sing the vowel sounds more properly, such that she did not confuse "Short a" for "Long A" and etc. helped a bit. "Ah Maze Ing Graess"...


--Mac

I worked with some early speech synthesizers in the 70s, and we had to program words with exaggerated phonemes along the lines Mac just described. "Hello" sounded like a fast, clipped "HEElow". We had to type "Hay-uh-LOW-uh" to make it sound realistic.

Meanwhile, I'm still trying to get sinsy to speak (or sing) to me....Hank

Some are definitely better than others, and the technology has continued to improve.

This is one example that demonstrates potential.

Although I think a lot of editing has been included in the samples to get the phonemes sounding somewhat natural. I doubt you can type a lyric and expect great results out of the box.

Voice synthesis speaking products are currently much better than those trying to include melodic phrasing.

In my view, Sinsy doesn't even rate.

A side benefit of Sinsy is that it is implemented by Music XML exported from BIAB. I'm hoping further development will be done in this area.

The merits of vocal synthesis are irrelevant in addressing the issue at hand. The fact is that the auto mode of generating synthesized vocals does not work at the moment.

I would be happy to hear mispronounced words with all kinds of artifacts. All I have been able to generate using the auto mode is the same "kih" syllable previously described. There is no hint of any kind of word pronunciation. I made a call to support last week and in real time he generated the same kind of file I have been getting, but sidestepped the issue by suggesting it might be a temporary problem with sinsy, a third party application.

On the other hand, I have been able to get more typical results using the manual process. Inserting the generated XML file into synsy and then importing the synthesized vocal wav file into RealBand or BIAB seems to work for some reason. I don't know what is different in the auto mode, but this could be a lot of fun and gets the creative juices flowing. I'm sure, like every other new feature, the kinks will be worked out, but we should not confuse what is being generated by the auto mode with the shortcomings of vocal synthesis.

Hank, you should be able to manually create a file from Sinsy. It sounds like you're still having difficulty doing that. Here's what to do:

1. Add lyrics to the song. It sounds like you've already done that, so I'll skip details on this step.

2. In the Lyrics menu, choose Vocal Synth- Manual Mode (Generate Vocal Audio from Lyrics & Melody).

3. You should get a popup message box with text along the lines of:

C:\bb\data\SRequest\Sound.XML is now generated.

Use your browser to browse to www.sinsy.jp, upload the xml file,
and press OK here when you have a file returned from
Sinsy.

At this point, BiaB is waiting for you to manually load the file onto the Sinsy website. Pay attention to the path the BiaB gave you. In my example, C:\bb\data\SRequest\Sound.XML is the location of the file to load into Sinsy.

4. Just to make sure everything is cool, open the folder (in my example, C:\bb\data\SRequest) and verify the file Sound.XML is there.

4a. (Optional) If you're really curious, you can double click it and have a look at it. It'll probably open up in your web browser. XML (eXtensible Markup Language) is related to HTML, and is a structured tagging language. You'll find the syllables of your song embedded between <text> tags. Close the web browser when you're done looking at the file.

5. Open up a web browser, and navigate to http://www.sinsy.jp.

6. In the upper right corner, you'll see English / Japanese. Click English to see the web page in English.

Edit: I forgot this critical step: Make sure the Language option is set to English!

7. There's a Browse... button on the web page for you to load your song into Sinsy. Click it.

8. A File Open dialog will appear. Navigate to the location where BiaB indicated the Sound.XML file was. In my example, it's in the folder C:\bb\data\SRequest. Navigate to that directory, choose Sound.XML and click Open to load the file into Sinsy.

9. Once you've selected the Sound.XML file, click the Send button on the Sinsy webpage to process your song.

10. The word Synthesizing will appear. Wait a few moments, and the words will change to Synthesis Results. There should also be a playback control.

10a. (Optional) Click the ">" button to hear the vocals. This verifies that Sinsy generated the vocals.

11. Now you need to copy the song from the webpage back to your computer. Right-click (i.e. click with the right mouse button, not the left button you usually use). A special context menu should pop up under your cursor with items such as:

Open
Open in new tab
Open in new window
Save target as...
Print target...
...

Choose the Save target as... option. (In Firefox, the option is Save link as...)

12. Your web browser will open a Save As dialog. By default, Sinsy gives the file some automatic name such as "20140120071413_32118.wav". You might want to shorten the name of the file, but it doesn't really matter. Save the file somewhere you'll be able to find it, like your Downloads folder or your Desktop.

12a. (Optional) Once you've saved the file, you might want to check it. Open the folder where you saved the .wav file, and double-click the file. Windows should open up the media player and play the file back for you. Close the media player when you're done listening.

14. All this time, BiaB has been waiting patiently. Go back to BiaB, and press OK on the message dialog.

15. BiaB will open a BB File open dialog so you can point it to the .wav file you just saved. Using the dialog, browse to where the file was saved, and select that file.

16. If all goes well, BiaB will load the file into the Audio slot. This should be indicated by the Audio slot now being yellow instead of white.

16a. I noticed that BiaB didn't mute the Melody slot. You might want to mute it, or the melody instrument will play back with the audio.

17. Press the Play button, and your song should play back with the audio track from Sinsy.

Did this work for you? If not, at what point did you run into problems?

David (Dcuny)
Thanks for taking the time to give this thorough instruction path. I'm not sure everything you mentioned is happening when I open manual mode but I will try it a little later today, and again, thanks....Hank

Strictly speaking, not all the steps are necessary. But if something is failing, you'll be able to identify where things go wrong.

Unless BiaB is going to adopt something like Vocaloid, I think integrating something like SoftVoice is the way to go. That way, BiaB isn't dependent on some third-party tool that they've got no control over.

Good luck!

Well, just finished my last trial with it & conclusion: Sinsy the Singing Cat. PS i messed with vocaloid a while back, especially trying harmony vocals ..... got the same sort of results, funny but just not useful. - F

Great review, my experience exactly. Of course I remember when BiaB sounded pretty bad (especially in hindsight) too. Wouldn't be surprised if this technology get's a lot more useful in the future.

For comparison, here's a demo done with Sinsy and BiaB: Words Like Smoke

Compare that to a demo using the Vocaloid AVANNA and BiaB: The Water Is Wide

I've tweaked the syllables a bit in them, but I think they're representative of the sort of sound you can get out of them. Sinsy has a heavy Asian accent - she a lot of trouble with r and th, for example.

They both come off as heavily retuned, because of the nature of the resynthesis. I've also had someone use the word "funny" when hearing it. Neither sounds like a native speaker.

But to see how far vocal synthesis has come, here's 1970's style formant synthesis (my own code): Loch Lomand

We explored Voice Synthesis some time ago, including having developed applications that use it.

The field is continually improving.

From the research that we undertook previously the Acapela Box product has potential, and also Ivona Reader .

Note - these are voice, not singing, but in terms of phonetics and level of clarity, they seem to be significantly better than the Sinsy product.

David (dcuny)

I am now able to navigate to the sound file but there is no XML extension attached to it. Vocal Synth follows the tune, singing six, six, six etc. and is also lagging by about 80 clicks ...Hank

Most voice synthesis programs use training databases to generate transitions between frames (slices of sound). This captures all sorts of prosody information, such as pitch and stress. The approach is generally referred to as HMM (Hidden Markov Model).

Sinsy takes the same sort of HMM as Ivona Reader does. The fact that she's got a heavy accent actually demonstrates the strength of the approach, since it captures the speaker accurately.

But to do this same approach with singing, you would need a huge database to capture all the transition sounds at all the different pitch intervals. So it's not really practical to create a database that covers all the transitions.

Additionally, there are timing constraints in singing - something that vocal synthesis doesn't deal with at all.

So getting synthetic singing to work with "state of the art" vocal synthesis turns out to be problematic, to say the least.

The extension is probably hidden, since Windows does that by default. So far, so good.



That's odd. Could you email me the XML file? I've sent you a PM with my email address.

Thanks David
I will work on that tomorrow and I really appreciate your help...Hank

David (dcuny)
I inadvertently shut down the forum before copying your e-mail address. Would you please send it again...Hank

David (dcuny)
I just listened to you works. I felt that the second sample was better than the first, but both are fantastic. The BIAB backing is very well done. I enjoyed your earlier work too, the male voice and harmony parts. Wish I had more computer savvy ...Hank

Hi, Hank.

You should be able to find the message by going to the menu at the top of the forum and picking My Stuff and Messages from the dropdown menu.

Thanks for the comments, but this is something that anyone could do with BiaB and a Vocaloid. I'm not sure what direction PG Music wants to go with vocal synthesis, but I think it's certainly an itch that some folk want scratched. Sinsy is certainly an inexpensive way to start.

The Sinsy code has actually been released, but I'm taking a vacation from coding for the moment. I might have a peek to see how they do the pitch shifting at some point, though.

That's actually been done quite often. It's called reggae.

(-:

But I side with the nay sayers and can't imagine why PG spent 5 minitues on this feature.

The issues go way beyond the technical ineptitude of voice synthesis to the very heart and soul of great singing which is
PHRASING!

Relatively few singers can phrase worth a darn and almost no non-singer musicians would have a clue. So "note based" vocal synthesis...even if the voice could be made to sound totally lifelike...would be doomed to sounding S...Q...U...A...R...E!

(-:

Jim

Thanks for the feedback, Rutabago.

I just played around with the Matsua-P voice - it's fairly good, and seems not to have the sort of noise problems that the Xiang-Ling voice has.

I think I may have given Hank bad instructions by not mentioning that English needed to be selected in manual mode. His Sound.XML file renders just fine.

"But consider that there are some folks who cannot speak or sing anymore, like our own good forum member Don, who just might be able to get enjoyment out of their songwriting and demos from the use of this feature."

Mac as Ive had health issues and hadnt been on here for some time I was taken back by what you said above..what has been the cause of Don not being able to sing or even speak? I absolutely agree with you for someone like Don this would be a wonderful piece of equipment,Thanks for posting it up Frankie

Sent you a PM, tributeman...


--Mac