Add --no-tts optional flag to disable XTTS

README.md

@@ -24,7 +24,8 @@ Download Automatic1111 here:
 Rough notes - I will polish the README later:
 
 ### XTTS
-- Current revision uses XTTS2 (uses the TTS Python library, lookup on coqui.ai) and it is REQUIRED (will make it as an option later)
+- Current revision uses XTTS2 (uses the TTS Python library, lookup on coqui.ai)
+ - if you do not want to use TTS pass --no-tts like so: 'python bot.py --no-tts'
 - XTTS2 uses some RAM/VRAM so bear that in mind
 - for setup I used windows gpu steps from https://github.com/daswer123/xtts-api-server (no need to clone xtts-api-server, just do the install steps from its README)
 - definitely need 'pip install pillow' and 'pip install TTS'
@@ -78,6 +79,7 @@ To run this bot:
 4. Install the requirements. I suggest using an Anaconda or Miniconda instance.
  ```pip install -r requirements.txt```
 5. Run the bot with `python bot.py`
+ - optionally with --no-tts flag
 
 Cheers!
 

bot.py

@@ -13,15 +13,17 @@
 import re
 import unittest
 import logging
+import argparse
+
 from typing import List
 from pydub import AudioSegment
 from PIL import Image
 
-# For xtts2 TTS
-import torch
-import torchaudio
-from TTS.tts.configs.xtts_config import XttsConfig
-from TTS.tts.models.xtts import Xtts
+# For xtts2 TTS (now imported conditionally at the bottom of the script)
+#import torch
+#import torchaudio
+#from TTS.tts.configs.xtts_config import XttsConfig
+#from TTS.tts.models.xtts import Xtts
 
 from aiohttp import ClientSession
 from aiohttp import ClientTimeout
@@ -34,6 +36,9 @@
 # Your API keys and tokens go here. Do not commit with these in place!
 discord_api_key = "PUT_YOUR_API_KEY_HERE"
 
+# Flag to disable TTS (XTTS) - use --no-tts when launching
+disable_tts_global = False
+
 intents = discord.Intents.all()
 intents.message_content = True
 
@@ -454,6 +459,9 @@ async def send_to_stable_diffusion_queue():
 
 # New function to handle TTS generation
 async def generate_tts(text):
+ if 'Xtts' not in globals():
+ return None
+
  await functions.write_to_log("Generating TTS for the given text...")
  out = tts_model.inference(
  text,
@@ -593,35 +601,39 @@ async def send_to_user_queue():
 
  # Do not do TTS for image gen responses
  if not reply["content"]["image"]:
- try:
- # After getting the dialogue, split it
- dialogue_parts = split_dialogue(reply["response"], 200)
-
- # Generate TTS audio for each part
- audio_parts = []
- for part in dialogue_parts:
- # better not send emojis to TTS
- part = strip_emoji(part)
- audio_path = await generate_tts(part)
- audio_parts.append(AudioSegment.from_wav(audio_path))
- os.remove(audio_path)
-
- # Create a silent audio segment of 0.5 seconds (500 milliseconds)
- silence = AudioSegment.silent(duration=800) 
-
- # Add the silent segment between each pair of audio segments
- combined_audio = sum(x for y in zip(audio_parts, [silence]*len(audio_parts)) for x in y)
-
- # Save the combined audio file
- md5hash = hashlib.md5(reply["response"].encode('utf-8'))
- md5hash_hex = md5hash.hexdigest()
- combined_audio_path = "tts_output_" + md5hash_hex + ".wav"
- combined_audio.export(combined_audio_path, format="wav")
-
- # Send the combined audio file
- audio_file = discord.File(combined_audio_path)
- except Exception as e:
- # TTS generation failed, skip TTS audio
+ if 'Xtts' in globals():
+ try:
+ # After getting the dialogue, split it
+ dialogue_parts = split_dialogue(reply["response"], 200)
+
+ # Generate TTS audio for each part
+ audio_parts = []
+ for part in dialogue_parts:
+ # better not send emojis to TTS
+ part = strip_emoji(part)
+ audio_path = await generate_tts(part)
+ audio_parts.append(AudioSegment.from_wav(audio_path))
+ os.remove(audio_path)
+
+ # Create a silent audio segment of 0.5 seconds (500 milliseconds)
+ silence = AudioSegment.silent(duration=800)
+
+ # Add the silent segment between each pair of audio segments
+ combined_audio = sum(x for y in zip(audio_parts, [silence]*len(audio_parts)) for x in y)
+
+ # Save the combined audio file
+ md5hash = hashlib.md5(reply["response"].encode('utf-8'))
+ md5hash_hex = md5hash.hexdigest()
+ combined_audio_path = "tts_output_" + md5hash_hex + ".wav"
+ combined_audio.export(combined_audio_path, format="wav")
+
+ # Send the combined audio file
+ audio_file = discord.File(combined_audio_path)
+ except Exception as e:
+ # TTS generation failed, skip TTS audio
+ audio_file = None
+ else:
+ # TTS disabled, skip TTS audio
  audio_file = None
 
  if not reply["content"]["channel"]:
@@ -654,16 +666,17 @@ async def on_ready():
  logging.basicConfig(level=logging.DEBUG)
 
  # Load TTS model
- await functions.write_to_log("Loading TTS model...")
- tts_config = XttsConfig()
- tts_config.load_json("xtts/config.json")
- tts_model = Xtts.init_from_config(tts_config)
- tts_model.load_checkpoint(tts_config, checkpoint_dir="./xtts/model", use_deepspeed=False)
- tts_model.cuda() 
-
- # Compute speaker latents
- await functions.write_to_log("Computing speaker latents...")
- gpt_cond_latent, speaker_embedding = tts_model.get_conditioning_latents(audio_path=[r"xtts/scarlett24000.wav"])
+ if 'Xtts' in globals():
+ await functions.write_to_log("Loading TTS model...")
+ tts_config = XttsConfig()
+ tts_config.load_json("xtts/config.json")
+ tts_model = Xtts.init_from_config(tts_config)
+ tts_model.load_checkpoint(tts_config, checkpoint_dir="./xtts/model", use_deepspeed=False)
+ tts_model.cuda()
+
+ # Compute speaker latents
+ await functions.write_to_log("Computing speaker latents...")
+ gpt_cond_latent, speaker_embedding = tts_model.get_conditioning_latents(audio_path=[r"xtts/scarlett24000.wav"])
 
  text_api = await functions.set_api("text-default.json")
  image_api = await functions.set_api("image-default.json")
@@ -862,6 +875,21 @@ async def parameter_select_callback(interaction):
  # Let the user know that their request has been completed
  await interaction.followup.send(interaction.user.name + " updated the bot's sampler parameters. " + api_check)
 
+parser = argparse.ArgumentParser(description='AI Discord bot, use --no-tts to disable XTTS and save some VRAM')
+parser.add_argument('--no-tts', action='store_true', help='Flag to disable TTS (XTTS)')
+args = parser.parse_args()
+disable_tts_global = args.no_tts
+
+if not disable_tts_global:
+ print("Running with XTTS (will eat some VRAM)")
+ print("Loading XTTS libraries...")
+ import torch
+ import torchaudio
+ from TTS.tts.configs.xtts_config import XttsConfig
+ from TTS.tts.models.xtts import Xtts
+ print("XTTS libraries imported.")
+else:
+ print("Running without XTTS - no speech audio will be generated")
 
 client.run(discord_api_key)
 #unittest.main()