SuttaCentral

SC Voice – the road to v1


#1

Following on from:

The marvellous work of @karl_lew on SuttaCentral voice has just seen the release of version 0.9.1 and he’s now on the home straight to see all his hard toil culminate in the delivery of v1 which will be linked on a new suttaplex card re-design on the main SuttaCentral site. Wonderfully exciting stuff!

As the above thread has gotten a bit long and unwieldy I figured I’d open a new one for any discussion pertaining to v1 development, which Karl aims to release in January. The main focus of work for this phase of the development is to improve Pali pronunciation with the phenomenal help Ang. @sabbamitta is giving and likewise, Bhante @sujato, particularly with info provided on nasals:

Additionally, there is an aim to see to small fixes and improvements that will give a little boost to the user experience and help to make SCV all the more enjoyable especially for new adopters, please do feel warmly welcome to give any such feedback here.

Beyond that, bigger items—eg. feature requests, accessibility research, etc—will all get shaken out in a post v1 head scratch (of course, you’re also warmly welcome to send me ideas of things to add to the possibilities list).

For now the plan looks like this:


SuttaCentral Voice Assistant
#2

And on it goes with bugs and pronunciation errors for the v1 release (or later):

  • Thig 13.5 only plays the introduction and then stops.
  • Thag 5.10 plays only the nice bell and then nothing at all. It shows “loading” for ages, but nothing happens.
    (But probably we shouldn’t focus too much attention on those; they’re not part of the segmented texts and will change in the foreseeable future anyway.)
  • SN 22.59 section 2 segment 8 ‘evaṃ me rūpaṃ hotu, evaṃ me rūpaṃ ahosī’ti.: the is pronounced as māha. It is clearly distinct from the following ahosī, so it’s not the two words blending into one another. The same occurs repeatedly in this text.
  • SN 22.59 section 2 segment 33 ‘etaṃ mama, esohamasmi, eso me attā’”ti?: The letter “e” occurs in etaṃ, and twice in eso, and to my understanding it should be pronounced in the same way in both cases—but it isn’t. etaṃ sounds just as expected, but in eso the “e” sounds rather dull (I don’t know the right words to describe these things).
  • SN 22.59 section 2 segment 45 nesohamasmi, na meso attā’ti: same problem with the “e”, correct in neso, and otherwise in meso. Amazing: In segments 46, 48, and 49 she gets it all right!
  • SN 22.59 section 2 segment 45, English “So you should truly see any kind of form at all—past, future, or present; internal or external; coarse or fine; inferior or superior; far or near: <em>all</em> form—with right understanding: ‘This is not mine, I am not this, this is not my self.’”—the html code is visible in the text.
    I’ve double checked on the main site, and there it’s not—but this may be one of the rare instances that indeed there is a word emphasized with italics in the text!
    The same occurs again in segment 49.
  • I’m again coming back to what Aditi does with double consonants and ñ, or especially ññ, like in viññāṇa. This in-between vowel, or half vowel, or how you want to call it (like in español) sounds very much like an extra sillyble, and this is somewhat problematic for chanting because it affects the rhythm of the text. When we learn Pali from Aditi we’ll get in trouble when we want to chant. With a word like viññāṇañca we fall completely out of rhythm!
    This may be difficult to fix, though, or even impossible; and I don’t think it’s a focus for v1 in any case.

Saṅkhārā sounds perfect now! Well done!! :anjal:

And to me it becomes more and more easy to discern the ṇ sound. :grin:

It’s fun all this, thank you so much! :heart:


#3

Maybe we simply shouldn’t bother at all about this. Amy doesn’t pronounce the code, and after all, this app is being designed for the blind. Visible, but not audible code is for a blind person just like a lint on the clothes: it doesn’t exist for them. Why bother a blind person with a lint brush when there is nothing to brush away?

@karl_lew, what do you think?


#4

I actually noticed that with another text and already added it to the to-do list, but these things are adjusted to how much work they take. If it’s a quick fix, why not, if not then, indeed, it should be defered.


#5

I see the HTML and I think it should be turned into a list so for the blind. Screenreaders understand list elements. :white_check_mark:


Here is a sound excerpt for Amy saying Jeta’s grove twice. The first is the current way. The second is the proposed way. For the second I simply doubled the vowel. This is not quite the same as a long vowel spoken in twice the time, but it is the best I can do. Shall we keep what is or go for the double e?

Yes. Aditi’s cadence is improper. I have noticed that as well. I think the ordained shall have to correct us as we chant as people. I see no way of correcting Aditi’s cadence. We would need a new, Pali voice.


I have copied all your other finds to the Release Plan

Thank you!


#6

Just to note, Thig and Thag are nearly ready for publication in segmented versions, I hope to have them done by the end of the year.


#7

This is magnificent! Suttacentral just gets more amazing with time.


#8

Thank you, Bhante.

I have added Thig/Thag content update to the Release Plan. Content update is currently a manual process so I have added content update automation to the Release Plan as well.


#9

One day I hope we may do Spanish segmented suttas! :smile:


#10

I completely forgot to answer that one last night. I’m probably just as unsure as you. Maybe, if it’s not much work, go for the double e; but if it takes you time it’s rather not worth it, I think.


#11

I have made the change to Amy’s voice for v1 but have left Raveena and Russell unmodified. Raveena was not quite as bad as Amy. However, Russell completely butchered the doubled e. At least now Amy will not sound like a Volkswagen commercial for a Jetta. :smiley:


#12

Sorry to just barge in on the convo here; not sure if this will be helpful, but what about a dipthong/gliding-vowel for ‘e’ of ‘a’+‘i’? If you can blend the transition so that the ‘a’ and ‘i’ are not pronounced individually but rather glided through.


#13

Here we have:

  1. single e
  2. double ee
  3. dipthong ei

What sounds best?


#14

That’s not even bad! Not quite as it should be, but as long as we don’t have any better…

Thanks, @SCMatt!


#15

Excellent. Fix updated as per @SCMatt’s suggestion. :pray:


#16

I’m wondering—do you think the same trick with the diphthong would also work for Gotama?


#17

Here is Amy with dipthongs:

  1. gotɐma
  2. gəʊtɐma
  3. gaʊtɐma

For reference, here is Aditi with her en-IN pronunciation:

  • got̪əmo

I hear #2 as slightly better than the current #1. What do you hear?


#18

Yes, agree. It’s definitely better.


#19

Fixed for v1.0.0.:white_check_mark: Thank you.

@Aminah, we could actually have another interim release as v0.9.2 with these corrections and with Bhante’s pronunciation guidelines if you wish. The pronunciation fixes are subtle and take time. The v1.0.0 release is drifting towards the middle/end of January and it will be a while before I can get to the large work on downloading playlist MP3’s for offline listening.


#20

Another proposal for a search phrase: first point