How Immersive Reader could be more inclusive - by speaking our language and with our accents

Scottish Gaelic voices

Currently, Read Aloud can't read Scottish Gaelic. The only Scottish Gaelic voice that exists (as far as I'm aware) is Ceitidh, from CereProc.

Microsoft Edge browser

If you open a web page in Microsoft Edge, Read Aloud offers you a range of voices including Irish and Welsh language, but not Scottish Gaelic. Note that these voices work over the internet - they're not installed on your computer.

Microsoft say they have more than 270 neural voices across 119 languages and variants and you can try them out on the Azure Text to Speech page - but a Scottish Gaelic voice is not available. Why Catalan, Galacian, Irish and Welsh but not Scottish Gaelic? Microsoft may be working on it - this blog post describes the new voice building process and suggests that voices in different languages can now be created more quickly than previously. And Mike Tholfson's January 2020 blog post suggests that Microsoft offers Gaelic Translation - although I can't see a Scottish Gaelic option in the Translate tool in Edge as yet.

Gaelic speakers and learners should be able to use Read Aloud to read Gaelic websites with a Gaelic voice. At the present time this isn't possible.

Microsoft Word 365 Desktop App

Microsoft also provides voices that you can download to your computer or device but again a Scottish Gaelic voice isn't available.

If you open a Gaelic Word file in the Word 365 app on your computer and you're connected to the internet, Read Aloud in Word 365 now reads it with a Microsoft male or female online voice. Previously, Read Aloud used the Microsoft voices installed on your computer. In my tests, it mostly read with an English voice, but once or twice it read a sentence in a different voice - I think it was Orla, the Microsoft Irish voice - but I couldn't get it to work reliably.

If you're not connected to the internet, Read Aloud uses one of Microsoft voices installed on your computer and complains that the language doesn't match a voice on your computer:

This is a bit confusing and also not true, if like me, you do actually have Ceitidh, CereProc's Scottish Gaelic voice, on your PC! We licence Ceitidh for the Scottish public sector and you can download the voice from our Scottish voice web site. But Read Aloud doesn't seem to recognise Ceitidh.

It gets more interesting: you can actually use the Speak Button in Word to read Gaelic text with Ceitidh and if you download the Scottish Gaelic Language Pack, the spellchecker can read with Ceitidh as well. These tools even seem to recognise that the text is Gaelic and they switch to Ceitidh automatically, which is pretty clever. The video below shows it working.

So the Office 365 Speak button and spellcheck can read with the Ceitidh Scottish Gaelic voice, and the software even seems to recognise the Gaelic language and it switches to use the Gaelic voice. The Windows Narrator screen reader is the same - it can speak with Ceitidh and other SAPI 5 voices as well as the Microsoft voices.

Why does Read Aloud not behave the same way?

Well, it seems there are two types of voices that can be installed on your Windows 10/11 device - Microsoft's own Windows voices, and 'SAPI 5' voices. Most text-to-speech tools on your Windows computer can use either type of voice. As we've seen,  Windows Narrator, the Speak button in Office 365, and the Office 365 spellchecker can all read with both Windows and SAPI 5 voices but for some reason Read Aloud can only speak with the Windows 10 voices and so you can't use Ceitidh or any of the other Scottish SAPI 5 voices.

Scottish accent English voices

Microsoft's new online computer voices are really clear and natural and there is now a very good range of languages and accents. For English language we have accents from: Australia, Canada, India, Ireland, Philippines, South Africa, UK (these are southern English accents - someone needs to explain to me what a 'UK' accent sounds like) and several US voices. No Scottish voices. 

As we've seen, you can also download more Microsoft voices to install on your computer and the full list as of 14th January 2022 is listed in Microsoft's Supported languages and voices. For English language we have voices for Australian, Canadian, 'Great Britain', India, Ireland, United States but no Scottish accents  - or Welsh, Irish, Geordie, Mancunian, Yorkshire, Liverpudlian, Cockney for that matter.

The Microsoft voices are all adult - there are no child voices.

Learners in Scotland should have one or more Scottish voices installed on their Windows computers in school. We have adult voices (Heather and Stuart); teenage voices (Isla and Callum) and child voices (Andrew and Mairi) and they can be downloaded from our Scottish Voice web site. The voices are all from CereProc.

Again though, Read Aloud is limited to using Microsoft's own English languages voices and can't speak with these SAPI 5 voices.

Does this matter though? What's wrong with reading or listening to digital text spoken with a southern English adult voice? Does a Scottish accent make a difference?

I believe it does.

In 2007 I made a proposal to Scottish Government to license Heather for the Scottish public sector and wrote

We believe there is a strong cultural and educational imperative for pupils in Scotland to be able to read and listen to Scottish curriculum resources spoken using a Scottish synthetic voice. 

Now I don't have any research findings to support this view, but I do have 12 years of experience working with learners who use text-to-speech in Scottish schools and I think it's extremely important that learners can use familiar accents and that younger children have the option to use a child voice, particularly when proof-reading their own typed text.

Solutions

There are a few possible ways forward that could help Scots and Gaelic learners (and people in other countries/cultures) be more included in the wonderful Immersive Reader world:

• Microsoft could develop and provide voices in a wider range of accents, ages and languages, including Gaelic.
• Microsoft could work with voice providers like CereProc, Acapela, Ivona and Nuance and support them to develop Windows 10 versions of their voices that work with Read Aloud.
• Microsoft could adapt Read Aloud so that it can speak with Windows SAPI 5 voices. I suspect that this is probably easiest and quickest - since the Speak Button and spellchecker can use SAPI 5 voices, it seems reasonable to assume that Read Aloud could be tweaked so it can also use SAPI 5 voices.

There's also an argument for inclusion - by opening up Read Aloud and Immersive Reader to other voice providers, users have more choice and greater diversity. Our Scottish voices are a case in point: if Microsoft's SAPI 5 voice system had not been open to third-party developers in 2007 we wouldn't have been able to develop and provide the Scottish or Gaelic voices for every Windows computer in Scotland and enable learners to use the voices with (almost) any text to speech program. The open SAPI 5 technology has enabled the development of a much richer and more diverse text-to-speech landscape and I think it's vital that Microsoft's text to speech systems continue to embrace as many languages and accents as possible.

It's not just an issue for Scots or Gaels either: someone from the North of England might want to use Jess; Welsh speakers might prefer the Welsh Ivona voices; children in the US or England may benefit from reading text with child voices from CereProc. Voice diversity is particularly important for people who use voice output communication aids and who want to create their own bespoke voice, like Lee Ridley (Lost Voice Guy).

Your voice is part of your cultural identity. Learners in Scotland who use Immersive Reader to address barriers to learning should have the option to use Scottish and Gaelic voices.

(This isn't just about Microsoft: next week we'll ask ourselves why you can't install third party voices on an iPad.)

Thanks for reading and please tweet, post, email etc (tag Microsoft @MicrosoftEDU) so that we can make our voices heard in more ways than one!