charliegerard / speak-extension

Chrome extension with speech-to-text functionality to make your presentations more accessible with live captions

Speak - Chrome extension

Speak is a Chrome extension using the Speech Recognition API to provide a speech-to-text interface as an overlay on a website. The main goal is to allow people to make their presentations more accessible for the deaf and hard-of-hearing.

Important note:

As any speech recognition system, it is not working perfectly but hopefully it can still be somehow useful and will probably improve over time.

Current status:

So far, seems to be working on:

• Reveal.js
• MDX-deck
• Spectacle

It does not work on Google Slides but they now have a caption functionality implemented.

How to install:

As a Chrome extension:

It has now been approved by the Google Chrome team so you can install it from the chrome web store.

Dev version:

• Download or clone this repo.
• Go to the Chrome extension interface.
• Activate Developer mode (top-right corner).
• Click on the "Load unpacked" button in the menu bar.
• Select the folder where you cloned the project.

You should now see the icon of the extension in your browser extension bar.

How to use:

• Click on the extension in your browser's menu bar.

• Select your spoken language.

• Click on the "Start" button.

• Allow access to your microphone if you haven't already.

• Detection:

  • If the website you are currently trying to use this extension on is blocking access to AudioCapture, you will see an error message displayed.
  • If you do not see an error message, you should be able to start speaking and see an overlay appear on the page with the caption.

• Click on the extension logo again in the menu bar to close the extension.

• To stop the tracking, open the extension and click on the "Stop" button.

If the Speech Recognition API does not recognise what you are trying to say, it will default to English (US).

Example:

Watch a video example

Languages supported:

(I found this list in a tutorial so not 100% sure it's accurate)

Languages	Region	Language code
Afrikaans	Default	af-ZA
Bahasa Indonesia	Default	id-ID
Bahasa Melayu	Default	ms-MY
Català	Default	ca-ES
Čeština	Default	cs-CZ
Dansk	Default	da-DK
Deutsch	Default	de-DE
English	Australia	en-AU
English	Canada	en-CA
English	India	en-IN
English	New Zealand	en-NZ
English	South Africa	en-ZA
English	United Kingdom	en-GB
English	United States	en-US
Español	Argentina	es-AR
Español	Bolivia	es-BO
Español	Chile	es-CL
Español	Colombia	es-CO
Español	Costa Rica	es-CR
Español	Ecuador	es-EC
Español	El Salvador	es-SV
Español	España	es-ES
Español	Estados Unidos	es-US
Español	Guatemala	es-GT
Español	Honduras	es-HN
Español	México	es-MX
Español	Nicaragua	es-NI
Español	Panamá	es-PA
Español	Paraguay	es-PY
Español	Perú	es-PE
Español	Puerto Rico	es-PR
Español	República Dominicana	es-DO
Español	Uruguay	es-UY
Español	Venezuela	es-VE
Euskara	Default	eu-ES
Filipino	Default	fil-PH
Français	Default	fr-FR
Galego	Default	gl-ES
Hrvatski	Default	hr_HR
IsiZulu	Default	zu-ZA
Íslenska	Default	is-IS
Italiano	Italia	it-IT
Lietuvių	Default	lt-LT
Magyar	Default	hu-HU
Nederlands	Default	nl-NL
Norsk bokmål	Default	nb-NO
Polski	Default	pl-PL
Português	Brasil	pt-BR
Português	Portugal	pt-PT
Română	Portugal	ro-RO
Slovenščina	Default	sl-SI
Slovenčina	Default	sk-SK
Suomi	Default	fi-FI
Svenska	Default	sv-SE
Tiếng Việt	Default	vi-VN
Türkçe	Default	tr-TR
Ελληνικά	Default	el-GR
български	Default	bg-BG
Pусский	Default	ru-RU
Српски	Default	sr-RS
Українська	Default	uk-UA
한국어	Default	ko-KR
中文	普通话 (**大陆)	cmn-Hans-CN
中文	普通话 (香港)	cmn-Hans-HK
中文	中文 (台灣)	cmn-Hant-TW
中文	粵語 (香港)	yue-Hant-HK
日本語	Default	ja-JP
हिन्दी	Default	hi-IN
ภาษาไทย	Default	th-TH

Tech stack:

• HTML/CSS/JS
• Speech Recognition API

Current known issues:

• The recognition is not always very accurate

---

To do:

• Fix UI so only displays a certain number of words so it doesn't take too much space on the screen.
• Support multi-language?
• Allow translation in other languages than english with Google translate? (Might be too slow...)
• Redesign