This project is a Discord-based Text-to-Speech (TTS) bot which allows users to play AI-generated speech within a voice channel using simple text commands.

This project leverages the ElevenLabs API:

• https://api.elevenlabs.io/docs
• https://github.com/elevenlabs/elevenlabs-python

Registration:

• Each user must first register their ElevenLabs API key with TTSBot by sending a direct message to TTSBot containing only their ElevenLabs API key (Create an ElevenLabs account at https://beta.elevenlabs.io/sign-up → Go to https://beta.elevenlabs.io/ → Click on top right profile picture → Profile → API Key)
• This is necessary since ElevenLabs only provides a limited number of characters per user for speech generation

TTS:
Speech is generated through the slash command /tts_normal

• The voices field lets the user select from the available ElevenLabs voices

• The unstable field lets the user choose whether to make the speech "normal" (false) or unstable/wacky/unpredictable (true)

• The msg field is where the user enters the actual content to be spoken by TTSBot

• Once each field is filled, sending the /tts_normal command will move TTSBot into the user's voice channel, where TTSBot will begin speaking

• After speaking, TTSBot will remain in the voice channel

Misc - Leave:
The user may evict TTSBot from its voice channel through the slash command /leave_voice

First Time Setup:

• Install basic Python requirements pip install -r requirements.txt
• Install ffmpeg to environment apt install ffmpeg
• Rename the file .env_example to .env after adding your bot's discord token and adding the name of your guild

To Run:

• python3 client.py

Short-Term ToDo:

• Reorganize Util and Speech classes for comprehensibility
• Modify FFmpegPCMAudio (in tts_normal, etc.) to use bitstream instead of wav
• Modify how user info (such as ElevenLabs keys) is stored
• Add logging https://discordpy.readthedocs.io/en/latest/logging.html
• Modify how ElevenLabs API keys are registered
• Implement robust error handling
• Conduct comprehensive unit tests
• Generally reorganize dependancy handling and code format to follow some kind of standard (lol)

Long-Term ToDo:

• Implement "One-shot" voice cloning using ElevenLabs API
• Transition from ElevenLabs to a locally-run model
• Implement "Deep" voice cloning with local training
• Control TTSBot through voice commands
• Distinguish between different voices (e.g. if Bill asks TTSBot for something, TTSBot will do that thing using Bill's stored info)