Today at 18:00 CEST (16:00 UTC) we are going to publish another new release of Voice.
This time the gap between our staging server and the current Voice production server is much bigger than it has ever been, therefore the release process can be expected to be a bit bumpy. It may even happen that the application will break down during the process.
We are standing close by with the fire extinguisher and will let you know when everything is back to work.
So now our beloved TTS app has been successfully updated—without any trouble, which was a bit unexpected. Voice 2.3 is up and running.
The most exciting news first: The Voice team has new additions to help us make Vocie even better in the future: Thank you very much, @Robbie and @Madhu, for supporting our team! By joining us you have just doubled the number of our work forces … + !
We are also very pleased that Voice has even more new language faces now: New web interfaces have been added for Czech and Sinhalese. Thanks to Venerable @sarana and Madhu!
Maybe the Russian translators @DZT and @DF_OWL would also like Voice to have a Russian web interface so that Russian speaking users can listen to their translations in Voice once they are published?
It might sound like this:
If you are interested, please let us know, and we’ll provide you the necessary file for the interface for translation (about 230 terms).
This release was supposed to be mainly about pronunciation, but as so often in life, thigs don’t always go exactly the way they have been planned. Much of the extra issues concern things behind the scenes that aren’t visible to the user, but without which Voice wouldn’t work as expected.
One thing is that the SuttaCentral main page is working hard on a new round of updates which also require some adaptations on the side of Voice. These include among other things an adaptation of the code in order to read Vinaya texts once Ajahn Brahmali’s new translations are released. SuttaCentral is also implementing a process for moving new translations from “unpublished” status to “published” status. The implementation hasn’t been fully completed yet, and so Voice is at the moment resting on an intermediate step and will have to do further steps in the next development cycle.
On Bhikkhu @Rusdevamitta’s request we have also added a new feature to speak single Pali words. If you click the new speech bubble button that’s left of the PLAY button you will be led to a little audio player that speaks the Pali word that you have typed into the search box. Much thanks to the Verable for this excellent idea!
The Voice search is now working faster due to using ripgrep instead of grep. It is also able to search for words that start with umlauts such as German Übung (training).
In case of server problems, there are now more than just one person who are able to start or stop and log in to the Voice servers.
Another important point is that we want to make Voice more user friendly for those who prefer reading over listening. So far it was possible to listen to Suttas either in root or translation language or in both, but the text shown for reading has always been in translation language only. Now you can read exactly the same languages that you select in the settings, i.e. monolingual or bilingual, and select between side-by-side (default) or line-by-line view (tick the respective box in settings). Just expand the little arrows that you see underneath the PLAY button to read the text. (For technical reasons, bilingual view is only available for segmented texts.)
The print output for Voice has also been improved so that the text is now nicely readable on white paper. (Tip: Don’t try it on Firefox ; Chrome works better. Other browsers have not been tested.)
The “settings” menu has been improved so that one section closes automatically when another one is opened.
We also created a new accompanying website with GitHub pages, About Voice. It hosts the content of the former Voice Wiki pages, but in a more accommodating and user friendly way. It has different sections for the different languages supported by Voice. You can access the language selection page by clicking the “i” icon in the top right corner of the Voice website.
So far the full content is available in English and German, and some parts have already been translated into Portuguese, Japanese and French. The German section also hosts German translations of some of the introductory texts and essays from the SuttaCentral main page (more to follow).
@Gabriel_L, @Kaz, @SeanL: If you wish you can review what you have already translated and adapt in accordance with the new outlook of About Voice. You can translate more pages as you wish or your time permits, and even add more material that may be relevant to your respective audience. If you like you can also customize the layout. Compare the German section which appears as a website on its own with its own name and layout that is related to the general layout of About Voice, but yet has its own flavor.
Of course all other translators who have translated the Voice web interface into their languages are also invited to translate the pages of About Voice if they wish. I can only tag in a limited number of people in one post, so will name them in a separate post below.
Karl has been working hard to fix the remaining pronunciation bugs for Aditi, the Pali voice. We feel it is important that the Pali you hear in Voice doesn’t teach the user a wrong pronunciation; or at least we want it to be as correct as technically possible. This time there were a lot of hiccups to fix!
Now that these issues for Pali pronunciation are fixed and SuttaCentral has done a thorough check of the integrity of its Pali text, we are preparing to build VSMs for Aditi. This will serve two purposes:
It should make audio play and download faster.
It is also supposed to reduce the costs for Voice in the long run. Currently Voice has monthly costs for TTS and server services of about USD30+, and we hope to reduce them by half, once all Pali VSMs are built.
We hope you keep enjoying Voice and are always happy to hear your feedback! If you want to talk to one of the Voice developers or admins, ask questions or bring something to our attention, you can call us: @devs-voice.
Please note also that some updates have been made to the language.ts files in relation with some of the new features that have been implemented in Voice. If you wish to update your respective translations that would be most valuable for the audience of your language.
That’s possible, and there are actually two ways to make it available.
Bhante Sujato has started recording both the Pali text and his English translation, and we currently have Suttas from SN1.1 until SN2.20. They are available on Voice just like other reading voices:
If you select “sujato” for either Pali or English, a default robot voice will jump in for the Suttas he hasn’t recorded yet.
To do this, however, requires a sophisticated process of audio editing. The recorded sound has to be segmented along with SC’s segment structure. @michaelh has done this for the Suttas we have so far, but I haven’t seen him around here for quite a while, and he might just be busy with other stuff.
Another option is to make recordings available via the “more” menu button of each sutta:
Voice requires one audio file (e.g. MP3) for each segment. This allows Voice to present the suttas spoken segment by segment along with the corresponding text. That is the technical requirement.
It is the logistics and production that gets complicated. Bhante Sujato recorded entire suttas that he recited and we have those recordings. @MichaelH is an audio production wizard and was able to put together a software pipeline that splits those files into individual segments by guessing where the breaks are. It works most of the time but sometimes the software guesses wrong.
A second alternative is for you to record one MP3 file for each segment of a sutta to ensure that the breaks are as you intend.
The third alternative is for you to record an entire sutta as a single MP3 file that Voice can link to. We would lose the text correspondence here, but your recording would be available and served entire.