server : add Speech Recognition & Synthesis to UI

[ElYaiko]

This PR adds a Speech Recognition & Synthesis to the UI (A simple voice mode).

Features added:
Talk button: Initiates speech-to-text.
Send after talk option: Sends the message after STT.
Voice option: Text-to-speech voice used for the bot.
Play/pause message: Play/pause message with selected TTS voice.
Play message after completition

Tested browsers:

• Chrome
• Firefox
• Safari

Tested OS:

• Windows
• macOS
• Linux (Requires additional packages for TTS: Guide)
• Android
• iOS

---

• I have read the contributing guidelines
• Self-reported review complexity:
  • Low
  • Medium
  • High

[ggerganov]

Nice! Just need to fix the trailing whitespaces (see the CI)

[ngxson]

LGTM. Small detail: I'd prefer to add a small message says "TTS and speech recognition are not provided by llama.cpp", so to be clear to user that the quality depends on their browser, not on llama.cpp or the model itself.

[ElYaiko]

@ngxson What do you think?

[ngxson]

Yes it's LGTM, we can merge once the CI pass

[jboero]

Wow I just saw this update. Kudos merging Whisper and TTS this is brilliant. Well done.