Dear Microsoft Edge team,
I would like to suggest a feature that could give Edge Read Aloud a meaningful advantage over other text-to-speech systems, while also improving accessibility for people who are blind or visually impaired.
Edge is already one of the best tools I have found for reading English text aloud. However, like most text-to-speech systems, it sometimes mispronounces words that have the same spelling but different pronunciations depending on context.
Examples include:
- live — /lɪv/ as in “I live here,” but /laɪv/ as in “a live performance”
- bow — /boʊ/ as in “bow and arrow,” but /baʊ/ as in “bow before the king”
- lead — /liːd/ as a verb, but /lɛd/ for the metal
- wind — /wɪnd/ for moving air, but /waɪnd/ as in “wind the clock”
AI and text-to-speech systems can try to determine the intended pronunciation from context, but they will never be completely reliable in every book, website, game, technical document, or unusual sentence.
A simple solution would be to allow HTML authors to specify the pronunciation directly using IPA, while keeping the normal written word visible:
<phoneme alphabet="ipa" ph="laɪv">live</phoneme>
or:
<phoneme alphabet="ipa" ph="baʊ">bow</phoneme>
Edge Read Aloud could follow a very simple rule:
If a phoneme pronunciation is provided, use it. If none is provided, use the normal text-to-speech pronunciation system.
This would allow website developers, ebook publishers, accessibility tools, and automated systems to guarantee the correct pronunciation of ambiguous words without changing what appears on the screen.
It could also work very well with AI. A tool could scan a book or webpage for words known to have multiple pronunciations, examine the surrounding sentence or paragraph, select the correct pronunciation, and add the IPA annotation to the HTML. Edge would then no longer have to guess when reading the document aloud.
This would be particularly valuable for blind and visually impaired users.
Microsoft already has excellent English speech synthesis. Adding a standardized way for webpages and ebooks to provide pronunciation information could turn that existing strength into a real competitive advantage for Edge.
P.S. Pronunciation support could also be the beginning of a richer set of optional speech annotations.
For example:
<phoneme alphabet="ipa" ph="buːm" voice="louder" voice="faster">boom</phoneme>
could tell Read Aloud not only how to pronounce boom, but also that it should be spoken more loudly and quickly.
Other annotations could control delivery without changing the visible text:
<phoneme voice="whisper">wait, get down</phoneme>
Future attributes could potentially describe things such as whispering, shouting, speed, emphasis, pauses, pitch, emotion, or character voices.
These should remain optional, with ordinary HTML continuing to work exactly as it does today. But for ebooks, interactive fiction, educational material, accessibility applications, and other spoken content, they could make Read Aloud much more expressive and accurate.
For starters I think pronunciation support alone would already be extremely useful, and it could provide a strong foundation for much richer accessible spoken content in the future.
Dear Microsoft Edge team,
I would like to suggest a feature that could give Edge Read Aloud a meaningful advantage over other text-to-speech systems, while also improving accessibility for people who are blind or visually impaired.
Edge is already one of the best tools I have found for reading English text aloud. However, like most text-to-speech systems, it sometimes mispronounces words that have the same spelling but different pronunciations depending on context.
Examples include:
AI and text-to-speech systems can try to determine the intended pronunciation from context, but they will never be completely reliable in every book, website, game, technical document, or unusual sentence.
A simple solution would be to allow HTML authors to specify the pronunciation directly using IPA, while keeping the normal written word visible:
or:
Edge Read Aloud could follow a very simple rule:
If a phoneme pronunciation is provided, use it. If none is provided, use the normal text-to-speech pronunciation system.
This would allow website developers, ebook publishers, accessibility tools, and automated systems to guarantee the correct pronunciation of ambiguous words without changing what appears on the screen.
It could also work very well with AI. A tool could scan a book or webpage for words known to have multiple pronunciations, examine the surrounding sentence or paragraph, select the correct pronunciation, and add the IPA annotation to the HTML. Edge would then no longer have to guess when reading the document aloud.
This would be particularly valuable for blind and visually impaired users.
Microsoft already has excellent English speech synthesis. Adding a standardized way for webpages and ebooks to provide pronunciation information could turn that existing strength into a real competitive advantage for Edge.
P.S. Pronunciation support could also be the beginning of a richer set of optional speech annotations.
For example:
could tell Read Aloud not only how to pronounce boom, but also that it should be spoken more loudly and quickly.
Other annotations could control delivery without changing the visible text:
Future attributes could potentially describe things such as whispering, shouting, speed, emphasis, pauses, pitch, emotion, or character voices.
These should remain optional, with ordinary HTML continuing to work exactly as it does today. But for ebooks, interactive fiction, educational material, accessibility applications, and other spoken content, they could make Read Aloud much more expressive and accurate.
For starters I think pronunciation support alone would already be extremely useful, and it could provide a strong foundation for much richer accessible spoken content in the future.