[Feature] Add automatic reading function?

[Lufffya]

Feature description

Can you add automatic reading function for chatgpt answers? As shown in the figure: I hope that after I say hello, there will be a reading button behind the chatgpt answer (Like those three). Click this button to automatically read the chatgpt answer(Hello! How can I assist you today?)

Motivation

I just think it can be beneficial to learning, because chatgpt originally supports multiple languages. When I have to ask questions in an unskilled language, reading can help me understand better and will be more interesting. Thanks!!!

Alternatives

No response

Additional context

No response

[dsernst]

Do you mean reading the response out loud, as audio?

[dsernst]

I have been playing with this, I was able to get some basic JS working to find the most recent reply and read it aloud using the built in SpeechSynthesizer API.

I added this to my ~/.chatgpt/main.js file:

// Say command: read text aloud
function say(text) {
  speechSynthesis.speak(new SpeechSynthesisUtterance(text))
}

// Find the most recent reply
function lastReply() {
  const replies = document.getElementsByClassName('prose')
  const last = replies[replies.length - 1]
  return last.textContent
}

// Register a sayLast function to the global window so we can call it later.
window.sayLast = () => say(lastReply())

// Register hotkey Cmd+S (Ctrl+S on Windows) to start/stop
document.addEventListener('keydown', function (event) {
  // Ctrl + S or Cmd + S hotkey
  if ((event.ctrlKey || event.metaKey) && event.key === 's') {
    if (speechSynthesis.speaking) {
      speechSynthesis.cancel()
    } else {
      sayLast()
    }

    event.preventDefault() // Prevent the default action of saving the page
  }
})
console.log('sayLast() 🔈 function installed, Cmd + S hotkey to activate')

Then in a conversation, I can press Cmd+S to have the most recent answer read aloud.

Previous: ~~Then in a conversation, I open the Dev Tools (Right click > Inspect Element), switch to the Console tab, and can invoke the sayLast() function to have ChatGPT's most recent reply read aloud.~~

---

Possible Improvements

The hotkey works pretty well for me, but some other options:

1. Add a button next to replies to have them read aloud, as OP originally suggested.
2. Add a toggleable preference to automatically read all replies aloud. Ideally this would be a little thing visible in the window, so it can be turned on/off easily.

[chuyueye]

Can voice engine pronounce bilingual text？Now I can only select on lingua in the setting

[lencx]

Can voice engine pronounce bilingual text？Now I can only select on lingua in the setting

It is a built-in browser API and does not seem to support multilingual synthesis.

[neoOpus]

it would be better also to use Edge TTS

[null15]

Can I use ElevenLabs for the TTS?

https://api.elevenlabs.io/docs#/voices/Get_voices_v1_voices_get

[LeonChow5929]

There is "No Data" when I choose Set Speech Language. Dose anyone could tell me way?

[OMGHWORK]

There is "No Data" when I choose Set Speech Language. Dose anyone could tell me way?

Workaround: open developer tools, execute speechSynthesis.getVoices() in the console and copy the voiceURI to the speech_lang parameter in chat.conf.json

[pablorq]

The workaround doesn't work on Ubuntu 23.04:

> speechSynthesis.getVoices()
> ReferenceError: Can't find variable: speechSynthesis

Any idea?

[Frank15140156026]

hello