SuttaCentral Voice v1.5 released!

Hi all the friends of SC Voice!

After some trouble—as usual—the next version of Voice, v1.5, has just been released.

Human voice recordings in Pali

We are very happy, and also a bit proud (although it’s probably others who should rightly be proud of this achievement), that we can present parts of the Samyutta Nikaya in Pali chanted by Bhante @sujato! :speaking_head:

You can now for example choose SN 1.19 on the main site of SuttaCentral, click on the little loud speaker icon in the right upper corner which will take you to Voice, and then you should select your settings as follows:

Bildschirmfoto%20vom%202019-06-24%2015-13-18

Click on Play, and Bhante will be right there in your living room!

The Suttas currently available with human voice are SN 1.1–59. More will come as Bhante’s recordings go on and @michaelh can upload them. In a subsequent release we will hopefully be able to do the same for the English recordings.

We wish to express our utmost thanks to both Bhante Sujato and Michael for making this awesome experience possible!!! :pray: And of course to Karl, without whom Voice wouldn’t even have taken it’s very first breath! :dash:

Facelifting for the Voice settings

You will find the Voice settings with a new outlook now, and they include the option to select Bhante Sujato’s voice for the Pali text instead of Aditi’s. If you select this, Voice will still fall back to Aditi in all cases where there is no Sujato recording available.

And the settings also have some utterly cute icons now… :robot: :speaking_head: You should enjoy them for that reason alone.

The realization of which has come to be by the joined forces of Aminah and Karl. :pray:

VSMs

We have also started to build “VSMs”. VSM means “Voice Sound Module”. A VSM is a unit of voice segments stored together, usually the voice segments of one Nikaya in one language spoken by one particular speaker, robot or human. So for example all the segments of the Khuddaka Nikaya in Pali spoken by Aditi form one VSM, another one would be the same content spoken by Sujato. (The parts of the Khuddaka Nikaya currently available in segmented form are Thig and Thag.)

Once a VSM has been built and pre-stored on a server, SuttaCentral Voice will then be able to draw directly from there when you play or download a Sutta of this VSM. This means the latency for Voice to create the sounds each time falls away. VSMs will make your listening experience swift and immediate, and you won’t have problems downloading the content.

Since the creation of the sounds is what Voice is billed for, we are building the VSMs gradually over time in order to make the best use of our free tier opportunities for the robot voices we use. We are first building the Pali Aditi VSMs because all known pronunciation bugs for Aditi have already been fixed. (Changing one phoneme in a VSM will require to re-build the whole thing!) We have started with the smaller VSMs and are gradually moving to the bigger ones. For now we have the Khuddaka and Digha Nikayas by Aditi. The next one to follow is the Majjhima Nikaya.

Building a VSM takes several hours. Voice’s I-don’t-know-what-enthusiastic-adjective-to-use-ish engineer Karl :star_struck: has made it possible that a person with virtually no knowledge of website development like me (well, meanwhile I don’t run away screaming any more when seeing an html document, :scream: but hearing the word “Regex” still makes me tremble and shiver a bit… :face_with_thermometer:)—exactly: Our awesome Karl has made it possible that any silly Voice admin can now create a VSM. So that’s what I did, and it was very fascinating to see how this happens! And I am looking forward to do more of this in the future… :grin:

Another big THANK YOU to Michael for providing us a server for the creation of the VSMs!


For all items of this release see here.


We would still like to thank all of you who contributed to the thread “How do you use SuttaCentral Voice?”. It is very encouraging for us to see that what we are working on is used and appreciated, and that people find it helpful; and your suggestions give us inspiration for future developments. :sunflower:

14 Likes

To all who have contributed to make this happen

Sadhu! Sadhu! Sadhu!

8 Likes

Great team work!
Congratulations to Bhante @sujato for achieving his music ambitions. Finally!
:smile:

5 Likes

Very sorry to say that this doesn’t quite work yet. :crying_cat_face:

We have still some homework to do and will inform you when you finally can expect the

listening experience.

5 Likes

Oh dear. Voice has crashed. Investigating…

2 Likes

Voice is back up but in a fragile state. Disk space is almost all gone. We will be monitoring closely and will try to find ways to free up disk space.

5 Likes

I am glad to announce that the Khuddaka and Digha Nikaya VSMs could be installed on the Voice production server. However, they may have to be re-built and re-installed in order to save disk space, so they may be out of function again for a short while. We’ll let you know.

4 Likes

So happy to announce that more of Bhante Sujato’s recordings have been uploaded: SN 1 is now available in full, and SN 2.1–20 (all in Pali).

Enjoy!

5 Likes

:man_cartwheeling: thank you, @MichaelH and Bhante Sujato. :tada:

6 Likes

Hmm I ran into a bit of a bug today (sorry if this isn’t the right place to post bugs):

Occasionally Amy would read the first part of a segment twice. On occasions where she would do so, she would then also cut off the last few words of the segment (presumably to ensure she still ended on time).

(In case it matters, this was where it happened)

4 Likes

That’s perfectly the right place, thank you! I’ll investigate the case and see what we can do… :grinning:

Yes, it’s very helpful to point to the relevant passage. (If it’s a longer sutta it would even be good to mention the exact segment.)

Could you please give me the segment number; it is not that short after all (like “sn36.11:2.8” for example)? You find it at the bottom of the sutta player. Thanks a lot! :pray:

3 Likes

Sadly, it doesn’t seem to be consistent :frowning: But it happened with the Pāli voice on segment 2.20 just now as I was trying to repro (but it didn’t on the English where it had before).

3 Likes

My internet here is a bit spotty. Could it be something like:

The JS thinks the first request failed or timed out and so it retries but then the the first audio starts anyway (just late), then the second request audio starts when it comes back (cutting off audio 1 and going back to the beginning) but before audio 2 finishes, audio 1 reaches its “I’m done” time-based callback and then cuts off audio 2 midsentence.

3 Likes

We should defer this question to @karl_lew. I have no idea what his code may think… but I am convinced it does think in some way! :robot: :laughing:

I have just listened to the entire sutta until sn36.11:2.21, and it was all good. So maybe it has something to do with spotty internet. :slightly_frowning_face:

3 Likes

For what it’s worth, chrome lets you simulate slow internet for testing such things. (My internet registers as “Regular 3g” on Google’s scale)

4 Likes

Cool! Will try that…

3 Likes

Listening to the entire sutta on Chrome with Regular 3G, I can’t reproduce the bug, unfortunately.

But :scream: ! My computer becomes just so slow when I open two browsers at the same time! (I usually use Firefox.) Anything I want to do, like just clicking on another tab in the browser or opening a file, takes ages!!! So yes, that was a slow testing. :grin:

3 Likes

Maybe just one more question, @Khemarato.bhikkhu: Does it still happen for you after we have played this sutta a few times? If no, it might also be a cache problem.

2 Likes

It just happened again on segment 1.4. Amy said, "Pleasant, painful and nePleasant, "

3 Likes
3 Likes