SuttaCentral Voice Assistant

Tags: #<Tag:0x00007fc45742b3a0>



And so shall it be. One v0.9.1 coming up this week.


Exception: e and o change to short sounds in syllables ending in consonants. They are then pronounced as in “get” and “ox.”

(From Abhayagiri chanting book)

I have read this and always been wondering about an example for this, and never come across any.

Until now: In MN 11 section 2 segemnt 75 (and some other segments later) there is the word bhonto which Aditi pronounces rather as a long syllable. I’m not entirely sure how to divide syllables, but I’d think it should be bhon-to, so the first syllable ending in a consonant. Did I find an example for this exception from a long o?

In MN 11 section 2 from segment 109 on there occurs a few times the phrase Kiṃnidānā kiṃsamudayā kiṃjātikā kiṃpabhavā?. In segments 109, 115, and 117 Aditi doesn’t manage to properly pronounce the kiṃ in kiṃsamudayā.


That’s a natural inference. I’m going to hesitate here and say that we let this one slide absent any conflicting sound-alike word. The reason for this is that I can detect particular letters, but not abstractions such as “two consonants together”. All those are handled in the black magic box of Aditi’s AI. With such fuzzy AI rules, a fix here may break something else there. Perhaps Hindi evolved away from Pali on this.

I’ve added this to fix. :slight_smile:


That’s right. As a general rule, when there is a consonant cluster, the first of the two consonants belongs to the preceding syllable, the second to the subsequent syllable.

Another example is met-tā.

Note that clusters with “h” as the letter following a plosive (th, ph, etc.) do not follow this rule, as “h” in these cases is not really a distinct letter, but merely a sign of aspiration. Thus dukkha breaks to duk-kha.

However, when “h” follows a non-plosive consonant, such as in taṇhā, it is indeed a genuine letter and keeps its place rhythmically: taṇ-hā.

That sounds reasonable.

Incidentally, on the old site we have a syllable-splitter, written by Blake in javascript. For some reason it didn’t make the cut on the new site, but we should bring it back! To use it go to any Pali page on the legacy site:

Then to the sidebar, “Controls”, and hit the “k-a” button.


Omigosh. I started reciting that in my head and hearing English.
Pali immersion is really working! It’s all rote, but wow. :open_mouth:
“So I have heard. At one time, the Buddha was staying in Savatthi, Jeta’s grove, Anathapindika’s monastery…”


We’re adding search examples and I have included key phrases found in these suttas per your suggestions:

  • don’t fear good deeds
  • two fools
  • know for yourselves

The thought is that students unfamiliar with the Dhamma or suttas would type in a phrase in their native language. These phrases will often bring up multiple suttas (a handful) and allow them to choose the one they like. For example, “don’t fear good deeds” brings up AN7.61-63, one of which is Mettasutta. Every time people click Examples, they get a random selection of three phrases:

It is easy to add phrases. Aminah is collecting a list. Ideally, each search phrase should result in at most a handful of suttas. This will let people see how the suttas are connected in common.


Ideas to add:

  • getting rid of resentment
  • darkness light


Maybe some similes too:

  • raft
  • water
  • fire
  • wood
  • finest lady


One question on the search results:

MN 79 The Shorter Discourse With Sakuludāyī — matches: 3 of 200

What does the “matches: 3 of 200” refer to?

For example it shows me 5 results, and each one has another number of matches: 8 of 32, 6 of 9, 4 of 279, 3 of 100, 3 of 200. What does it mean?


Amy: The same problem as for “Jeta” also occurs for “Vesālī”.

But sometimes I just think these are really trifling things, given the overall good quality of the recitation, both in Pali and in English!


MN 12 section 2 segment 31:

Again, Aditi is getting a bit overexcited for a certain passage, anuttaraṃ vā cittaṃ anuttaraṃ cittanti pajānāti; samāhitaṃ vā cittaṃ samāhitaṃ cittanti pajānāti, the voice repeatedly changing pitch and failing to pronounce properly.


The same again at MN 12 section 2 segment 50, for the last bit: paraṃ maraṇā sugatiṃ saggaṃ lokaṃ upapannā’ti. Iti dibbena cakkhunā visuddhena atikkantamānusakena satte passati cavamāne upapajjamāne hīne paṇīte suvaṇṇe dubbaṇṇe sugate duggate yathākammūpage satte pajānāti. But already from the beginning of the segment she seems to quite speed up her voice!

Actually, when comparing it to “normal” segments, in these “fast” segments Aditi seems to treat whole sentences as one word and doesn’t recognise when a new word starts. This is throughout the whole segment, and some passages she just can’t manage at all any more, and her voice just starts somersaulting.

When this happens it’s (so far) always in long segments.

Again in segment 51


These are problematic in their ubiquity. For example there is wood and Great Wood. Do we have key phrases using these words?

Scoring relevance is complicated.. Let us know how to improve the documentation.

Amy is quite definitely the tourist in Pali land, we will have to cringe and wince. I thought of splicing in Aditi words but it would just feel like, well, Frankenstein.

Ah! Perfect. The warbling pitch is a sign of AI in pain. And the speeding up is a concern. In some cases I’ve had to remove the speed bumps just to fit the Pali into the character limit for AWS Polly. We’re on a strict budget and this may be the problem. I’ll look into it.

Thank you. :slight_smile:


Your explanation is clear and understandable to me. Thank you!

Will think about it.

Well, let’s cringe and wince a bit, rather than having Frankenstein. :scream:

So I don’t need to report such instances in the future?

There will certainly be more of these. Should they be reported individually?


I would compare Amy/Russell/Raveena to the voicing of your computer’s screenreader, which quite often will completely butcher the Pali. At least Amy/Russell/Raveena do try their best. When I use ChromeVox on my Chromebook, the Pali is quite unintelligible and sounds like a befuddled tourist with speech disability. At least Amy/Russell/Raveena have some idea of pronunciation and are somewhat intelligible. With the secret goal to teach everybody Pali by immersion (i.e.,. jetavane vs. Jeta’s Grove) I’m hoping they will learn to speak Pali as it should be and then complain that Amy/Russell/Raveena aren’t correct. That would be a big win if everybody complained about Amy/Russell/Raveena. It would be a big win because they would therefore know how to say it properly. Was that byzantine? :stuck_out_tongue_closed_eyes:

Regarding the warbling pitch, I would be sensitive to your time and not impose. You have given me quite a lot to fix and would be spending time rehearing reported mistakes. The new release will be out Dec 3-7 and will have as many fixes as I can muster. You’ll be able to verify the fixes then and continue the Great Hunt for the Gobbledy Gobbles.


Hi @karl_lew
I am having an issue when using the “Play All” option using SVA on my Android phone / Chrome browser.
The player g
Gets stuck in between sutras and when it does so it says “(no translation available)”.
I then have to turn on the screen, open the browser and hit next to force it into the next sutta
I am trying to listen to a whole chapter of AN in one go, e.g. AN7
Thanks in advance for your attention and help if it is something I am doing wrong on my end of things


I think that’s already been dealt with and the fix will enjoyable with the coming release.


How is the letter combination “ae” to be pronounced in Pali? As one sound or as two sounds?

In MN 12 section 2 segment 214 occurs the word naehibhaddantiko, and Aditi pronounces it as one sound, similar to German ä. Is this correct, Bhante @sujato?


There’s technically no vowel combinations in Pali, so things like this are really just a spacing issue. It would be more clearly spelled na ehibhaddantiko. In other words, each vowel is pronounced distinctly.


Fixed for v0.9.1. Thank you!