I’d like to download all DN suttas in Pali for offline usage from sc voice. But getting a “too long” error from sc voice for dn16.
I understand that there might be some capacity related reasons for that. But may be its possible that admin or developer will generate pali ogg file for dn16 and attach it somewhere? E.g. https://github.com/sc-voice/sc-voice/wiki/Audio-dn16. So not only me but anyone who is looking for pali audio of dn16 will have a chance to get it.
@sujato could you please let me know whom i can ask about it.
Ps By the way sc voice is only site (as far as I know) where it’s possible to get normal voice for pali without singing, chanting etc. Which is super cool and very important. Thank you so much for that opportunity to hear words, not melodies
Are you interested in Pali only, or in bilingual? Pali only might be easier.
I cannot provide the file because my internet is too slow and I get already time out errors for Suttas shorter than DN 16. But I will ask @karl_lew who has better internet. If he can’t download it, I guess nobody can.
In any case it is not possible to make changes to the Voice server. Voice lives peacefully and happily on its own server with its outdated software, waiting for its follow-up application to get ready (which may hopefully have better offline capabilities). So for the time being we have to live with what we are able to download from Voice the way it is.
It’s good that you need only Pali, for any updates on translations also can’t be kept up by Voice any more. And there have recently been quite a number of updates to the English DN translations!
There is hope that Karl can download DN 16 Pali only. He can for example download DN 33 (the second longest Sutta) in bilingual mode, while I can’t.
Voice software is currently frozen. What’s happened is that Voice has a dependency on Vue2, which is deprecated and Voice is not compatible with Vue3, which is the successor to Vue2.
The good news is that we’re working on a solution to this vexing problem with EBT-Vue3. In fact you can listen to DN16 using EBT-Vue3.
EBT-Vue3 differs from Voice in that it will cache audio on your local device. Because of this, we can now walk around freely listening to suttas on a phone. With Voice we used to have to download suttas as audio files, which can be VERY cumbersome, especially for longer suttas such as DN16. In other words, EBT-Vue3 brings audio for each segment down from the cloud and stores them locally without need for a separate download. The user experience should be much friendlier and versatile. Let us know how that works for you!
p.s., EBT-Vue3 is still under development and works best with Chrome. Also, EBT-Vue3 uses the same unadorned voices as Voice, so it will be quite familiar to you.
But getting Audio Error
audio.createAudioBuffer() invalid arrayBuffer
I’m using chrome mobile.
@karl_lew just to clarify, EBT-Vue3 will not have option to download audio file and user will need to open the page and listen sutta from it, it is correct?
By the way do you have plans to make a males voice? I’d like to imagine that listening for Tathagata or Sariputta or Moggallāna, but now all suttas are given by Dhammadina only
As a possible workaround, I found that changing the audio bell tends to prod the audio library properly. There is apparently some trickiness involved with getting audio to start in response to a user action for the first time. BTW I have an iPhone. Are you using Android (I don’t have an Android phone for testing)?
We have Bhante Sujato’s own voice for selected suttas such as SN1.1. Bhante has a long-term plan to record more, but he’s quite busy. And we do have male voices for translated narration. We have an upcoming need to update our AWS SDK to a newer version, so perhaps we can look at a male Pali voice once that update is complete.
Great! I can also use it on my Android phone without problems.
I think that depends on the experience we all—including users like you—make with the browser caching. If that fulfills the use cases sufficiently there is no need to invest the extra work.
Therefore it is so wonderful if you can try it out now and give us feedback.
Thanks for verifying the workaround. I will still need to fix this properly so that we don’t need the workaround.
Indeed. I no longer use the download capability myself since it’s so much nicer to just point at a segment now. Let us know how the download omission affects you as you try the new interface.
Downloading is combinatoric–users could choose arbitrary suttas spoken by arbitrary voices in arbitrary order. Therefore, downloading burdened the server with both CPU and storage concerns, so I’d prefer to deprecate that functionality and use server capacity for supporting more languages. It might (?) be possible, however to “glue together a single file from fully streamed, locally cached audio”, but that’s a rather large endeavor.
All Hindi bilingual voices usable for Pali are currently…female.
We have customized the voices for Pali, and Hindi has the best phoneme correspondence. Other languages with male voices garble the Pali unsuitably.
Id love to try. My main case is sutta listening in Pali in the forest. Or remote temples that usually are lacking internet connection. So to have a solid guaranty that ill be able to listen to sutta sound file is the nice option. And the perfect one is:
I use this one for listening in English and Russian. But of course Pali is missing. Ideally if there was a offline tts package downloadable as a other tts packages from play market or alternatives it did everything that is needed. But its a dream for now
Of course @karl_lew its much more convenient and staying in the city id love to have this experience. But usually its couple of days. And going back to the forest.
I tried to turn of internet on my mobile. And next paragraph got the “failed to fetch” error.
How can i cache the sutta before switching internet off?
Hi again @karl_lew, wonderful work on EBT-Vue3 sadhu sadhu and lovely to see you here again!
I’ve been following some developments in voice TTSes and there is a male Hindi voice in Azure cloud. Microsoft TTSes are considered by many to be state of the art now, and I checked that they can be configured with SSML too, in case you may want to try these out and it’s at all useful/feasible.
Lastly and most importantly there is some stuff I wanted to ask about remaining things for SC-Voice, and if it’s still using some things set up previously - may I DM you about some of the above?
But for my case with forest it looks like a scary task. So e.g. i need to play all dn suttas and itll take more then 24 hours. While downloading all files would take 1.5hr or so. But may be my usecase is unique and close to noone listens to suttas in offline environments. And anyway ive already downloaded all suttas that needed for many hours of the offline time.
So may be at least cache option can be a nice solution if you plan to avoid the downloadable file functionality? E.g. “cache sutta” so itll be ready for offline in minutes, not in hours.
And as far as i unsertand if user will clean the browser or reinstall it everything will be erased. So he will need to relisten all what he needs with internet again.
I did not know about Azure cloud voices. We won’t have time to look into Azure in 2023–our focus is on converting Dhammaregen and that has been like swimming in molasses with the Vue2->Vue3 transition and the redesign of Voice.
Quite amazing fidelity!
BTW, It’s interesting that VITS does great at the phoneme scale but struggles a bit with the narrative rhythmic and dynamic arc that frames the semantics of phrases such as"chopping and cutting" when compared with the ground truth. Despite that, it’s perfectly intelligible. Technology has moved so quickly!
Sure! Please CC Ayya @sabbamitta as well so we can assess and prioritize.
Well put. You are wise to be cautious. EBT-Vue3 is currently designed for intermittent/slow internet use and is not robust enough to go 100% offline. Looks like we’ll need to implement download in EBT-Vue3. Starlink might be an option, but it’s not as robust as MP3 files in the event of a solar storm.
Part of my hesitancy to implement download was that it monopolizes server CPU and disk for those minutes of download. The EBT-Vue3 streaming is much more scalable since the server impact is stretched out over the listening time. We still have the download API on the EBT-Vue3 server, so we’ll just need to add new UI for downloading as well as a gentle request to not overwhelm the server.
Quite right. We use web local storage for audio chunks. These are more ephemeral than MP3 files.
For now, you’ll need to listen to DN16 “in town” but all the other suttas should be downloadable individually or in combination as a playlist from Voice for your peaceful forest meditations.
Oh, BTW, I fixed the non-playing audio on mobile devices. Mobile devices have precautions against auto-playing sound and require special code to enable audio. Desktop computers are not affected.
