What is the alternative for the Web Speech API?

Submitted by Jeroen van Dinther on Sat, 03/07/2026 - 15:32
Forums: 
General discussion

For years Chrome has been the browser we have relied upon for the "Speak The Words" content type.
Since Chrome 139 (August 2025), however, Chrome has been moving towards on-device speech recognition. In the latest release, Chrome 160, even the Web Speech API demo at https://www.google.com/intl/en/chrome/demos/speech.html is no longer working on my Windows 10 laptop. When pressing the microphone button on the page, it appears to be listening but does not process any microphone input.
It does still work in Edge, by the way, but that has a different downside, since it adds punctuation.
Support on iOS for iPhone and iPad was never there to begin with, but ironically it does still work in Chrome 160 on an Apple laptop.

From previous discussions about issues with the "Speak The Words" and "Speak The Words Set" content types, I know it uses the Annyang library, that relies on the Web Speech API. The Annyang library appears to be losing ground and has not seen any updates for a long time.

What are the alternatives?
Could H5P developers drop the Annyang library entirely in favor of a native implementation and migrate to a modern web speech listener?
Or are there still ways to create a patch that fixes the current issues?

 

Content types: 
speak the words
Speak the words Set
wordpress
otacke's picture

otacke

Sat, 03/07/2026 - 19:24

Hi!

I am not entirely sure if I understand your question.

The WebSpeech API is an abstract API. It only defines how to talk to a "speech to text" service and what the answer from this service is expected to be. It does not define how that service is implemented. That's up to the browser to decide - and can be changed as needed. It could

  1. use an external server that does the heavy lifting, or
  2. let the device do the job directly, or
  3. record the spoken word on tape, send the to a human transcriber and then send the result back, or
  4. ... 

All that would not concern the developer using the interface. Their implementation would still continue to work without having to worry how the speech-to-text conversion actually works or if it remains done the same way over time. Annyang may be an unnecessary layer in between, but that does not change the interface nature of the WebSpeechAPI.

As far as I know, there is no other open standard API, no other open service that you could use for free + H5P lacks support for transparently using external services anyway (https://h5p.org/node/1454313).

Cheers,

Oliver