Could someone please assist me in locating an audio file of Bhikkhu Patimokkha in Pali, presented in spoken form rather than chanting? Even a generated TTS file, similar to what’s voice.suttacentral.net, would be greatly appreciated, although I haven’t found it there.
PS: Venerables @sujato and @sabbamitta, perhaps you could provide guidance on where I might find such a file? I’ve tried searching on Google and YouTube, but all I’ve come across so far are chantings.
No, very sorry, you won’t find any Vinaya text in Voice currently, and probably for the foreseeable future. The software so far only allows for Sutta texts.
We might address that at some point, but there is still a long way. Currently @karl_lew, our developer, is working on a new version of Voice, as the old one has outdated software that isn’t supported any more—so no change can be done to the old version of Voice. Getting the new version ready is much more urgent, for otherwise Voice may die completely.
By the way will it be possible to generate custom audio with the new voice.sc?
It sounds like a nice feature. E.g. if a person wants to listen to sutta without …pe… he could at least manually create the text without …pe… and “paste is somewhere” in the interface and listen. But if there will be the same logic as in the old Voice. It’ll be impossible
As well as listening to vinaya texts will require extra development work. Rather then just giving the end user such option to generate any audio that user needs.
No, I am sorry, Voice will only be able to speak what is in the texts belonging to SuttaCentral.
This is in principle possible with the tts servise that we use, AWS Polly. But you have to create your own user account (which requires not only a username and a password, but also your credit card details).
There you can use the Hindi voice Aditi (on which our Pali voice is based) and paste your text. You can only paste small portions of text at a time, so for the Patimokkha you have to create several successive audio portions. Also, the pronunciation will not be as good as what we have—it took about a year’s work to fine tune Aditi to become a good Pali voice.
It’s all not as easy as one thinks, unfortunately …
A while ago several of the software frameworks Voice is using had earthquake-like upgrades. Among those are node, npm, Vue/Vuetify, NUXT, and perhaps a few more. The new versions of ALL of them are no longer compatible with Voice software.
As Karl’s machine has the latest versions of them all, he can no longer run Voice on his computer, already for many months. This means that when something breaks, there is no way to fix it—which is why we don’t dare to touch Voice software any more. Voice is just living peacefully on its own server, and we hope it won’t die before the new version is ready.
As we had already been experimenting with EBT-Sites such as Dhammaregen, we are now incorporating our experiences from there into the new version.
Compared to the old Voice, the main changes will be:
There is no longer a separate audio player. Audio is just played in the segment, which makes it much easier to navigate inside a Sutta and select what you want to listen to.
You can choose between dark and light theme.
Audio download isn’t implemented yet, but is in the long-term planning. However, with browser caching, offline use or use in a place with week internet is much easier; though not totally offline—you need to get your audio from the internet at least once.
Currently we are busy with bug fixing. That means Karl fixes the bugs, while I am building up the new Dhammaregen and thereby find them!
Big companies with many developers can do many complicated things. Voice has one developer who is able to work one hour per day, so our possibilities are a bit more limited.
Can you give an example of an app comparable to Voice where the user can just paste any text and get an audio output? I am not aware of any, and I don’t think you’ll find a tts voice who is able to pronounce Pali as Aditi does.
It’s on our list of tasks. Downloaded audio handles the all important “no-internet” use case. I hadn’t implemented it yet again in EBT-Vue3 because I was unhappy with the user experience of that feature and wanted something that permitted reading along with the audio. Downloaded audio in Voice did not support that.
One note however. If you play through an entire sutta on EBT-Vue3, then the audio should be on your mobile device, in which case it should be playable, provided you don’t refresh anything. That last bit about “don’t refresh anything” is horribly fragile, and I haven’t figured out how to handle that use case. I’m unhappy with the state of that current implementation, but it might work.
Well, I suppose it could be possible for one to write a custom translation using Bilara, then that text would become available for publication. However, SC has strict protocol for publishing text. I think that requires at least two reviewers and there may be other constraints.
My way of dealing with the “…pe…” is to simply take another breath and keep on with what I’m doing.
Our UI framework went through a major rev change. Vue2 became Vue3 and Vue3 was not happy with Voice, so I had to…rewrite it. In addition, the architecture was monolithic and not scalable. Caching all the audio for multiple translators, narrators and languages on a single server swamps that server. We needed to decouple the front and back ends so that we can scale intelligently. For example, Dhammaregen only needs German, so it can have its own server and not have to worry about Portuguese.
Since a rewrite was required, it made sense to address serious UI issues with Voice that made it difficult to use. For example we wanted the user to be able to select a segment and just speak that segment for an interactive play-through of the audio.
Another concept we wanted to support was three-column output. For this we introduced a notion of a “reference translation”. The beauty of the segmented architecture is that it allows us to do a three-way comparison with, for example, English/German/Pali. Ayya Sabbamitta and Ayya Vimala recently used that capability at a retreat to bridge across the needs of attendees with a sum total of three languages in the room.
Indeed. Such apps don’t handle Pali well. The Voice narrators have special code to support spoken Pali.
I would say it’s more common to have user generated input, rather then predefined content in tts apps.
By the way some e.g. t2s provides user with option to set custom spelling rules (unfortunately not enough to make retroflex and some pali specific cases).
Ideal case - what is needed globally (in general, not in the scope of new voice development) is the tts package that can be used as default tts in OS (android, ios etc) Samsung tts Google tts or similar for desktop OS and web apps.
In this case user will choose interface and app to use on his own and developer/s will not need to spend time on ui and backend. Only working on the engine itself.
Thank you, @dhammagift. I think the approach of the apps that you mention may be different from what we try to do with Voice. I know for example of browser extensions that they can’t handle the Pali, nor can normal screen readers for assisted users. This was the main reason why Voice was developed in the first place. It is supposed to be a tts that is specifically designed to read Buddhist texts.
Probably for other languages one of these apps would serve you well, but as you want the Pali, there are limitations.
Thanks so much Karl and Sabbamitta for the explanation. We’ve rewritten SC back end a few times, and the front end how many, 1, 2, 3, 4, 5 times I think, hopefully this last one will stay! So far Lit has been very stable on the front end, it is built close to browser APIs so I’m hoping any upgrades won’t break anything major.
Currently rewriting Bilara both back and front. The good part is, you can learn from your mistakes! We’re looking at using HTMX for the front end.