Writing to share an update on automated AI machine translation for sacrament and other church meetings. I am also sharing a PDF of a broader "Using Accessibility Technology" training I provided at our Fall 2024 stake conference. The training deck covers doctrine (briefly), assisted listening devices, human interpretation devices, Church policy on use of artificial intelligence, and machine translation. Disclaimer: some self-promotion incoming, but I hope you will agree it is for a good cause!
Our stake has had great need for translation in several languages including Spanish, Portuguese, Chinese, and Haitian Creole. While human interpretation remains the gold standard, there are often occasions where either interpreters are unavailable or interpretation equipment coverage is inadequate. One of our wards supports simultaneous interpretation into three languages, and while possible, it takes a lot of coordination! (Kudos to the ward leaders, members, and missionaries supporting in our stake and elsewhere).
Previous threads have covered several possible AI machine translation solutions, including using Zoom add-ons, Microsoft Translator, translation earbuds like TimeKettle X1, or even Microsoft PowerPoint in captioning mode [1, 2] (thank you, wgotzmann, dave.swartz, markww, and others).
Among the established commercial options, our stake has had the most success with Microsoft Translator. The app is free and available for Android, iPhone, and Kindle devices (desktop version discontinued). See slides 24-25 in the attached training for a visual guide and recommendations, including a custom CSS browser hack to improve caption visibility.
Microsoft Translator works well most of the time. However, there are several pain points and hiccups:
- Caption visibility: By default, captions occupy a small portion of screen real estate. While the mentioned CSS browser hack helps, setting this up for each display device is a little tedious.
- Meeting code exchange: There is currently no way to reserve static meeting links, so a new five-letter code must be generated each time a meeting is initiated. This alone would not be a deal breaker. However, ...
- Disconnects: If the client disconnects, they must rejoin with a new name (mildly inconvenient). If the listening device disconnects, the entire session must be restarted with a new code (not fun!)
Now, onto the exciting part...
---------------------------------------------------
I never thought I would be here. But, after hitting dead-ends struggling to resolve the issues above followed by several late nights working on this side project, I am excited to share something new—dedicated hardware and software that provides nearly plug-and-play automated text transcription, text translation, and spoken translation via the local meetinghouse WiFi. We call it ChapelTranslate.
Since October 2024, we have deployed the device to all five buildings across our stake. Aside from an issue with an audio cable getting unplugged, it has been working great! The service starts and stops automatically as needed and is available via a static link or by scanning a QR code in the hymnals (thanks to a member for this tip).
Okay. So.
Why am I sharing now?
- To gauge interest.
- To recruit beta testers!
Thank you for reading! I hope the above resources help engage all of our friends and members in the Gospel, regardless of hearing impairments and language barriers.
Br Kenny Aycock
Stake Technology Specialist
Dover Delaware Stake
Sapidus Labs LLC
info@chapeltranslate.com
---------------------------------------------------
Accessibility Training (Fall 2024)
See attached, especially slides 22-25.
Relevant Threads:
Thread 1
Thread 2