mirror of
https://github.com/SillyTavern/SillyTavern-Extras.git
synced 2026-04-24 16:39:06 +00:00
Fix coqui TTS module
This commit is contained in:
Cohee
@@ -17,3 +17,4 @@ silero-api-server
|
||||
chromadb
|
||||
sentence_transformers
|
||||
edge-tts
|
||||
TTS
|
||||
@@ -16,4 +16,5 @@ diffusers==0.16.1
|
||||
silero-api-server
|
||||
chromadb
|
||||
sentence_transformers
|
||||
edge-tts
|
||||
edge-tts
|
||||
TTS
|
||||
@@ -17,3 +17,4 @@ accelerate
|
||||
chromadb
|
||||
sentence_transformers
|
||||
edge-tts
|
||||
TTS
|
||||
62
server.py
62
server.py
@@ -82,7 +82,7 @@ parser.add_argument('--chroma-persist', help="ChromaDB persistence", default=Tru
|
||||
parser.add_argument(
|
||||
"--secure", action="store_true", help="Enforces the use of an API key"
|
||||
)
|
||||
|
||||
parser.add_argument("--coqui-gpu", action="store_false", help="Run the voice models on the GPU (CPU is default)")
|
||||
sd_group = parser.add_mutually_exclusive_group()
|
||||
|
||||
local_sd = sd_group.add_argument_group("sd-local")
|
||||
@@ -248,6 +248,11 @@ if "silero-tts" in modules:
|
||||
tts_service.update_sample_text(SILERO_SAMPLE_TEXT)
|
||||
tts_service.generate_samples()
|
||||
|
||||
if "coqui-tts" in modules:
|
||||
mode = "CPU" if args.coqui_gpu else "GPU"
|
||||
print("Initializing Coqui TTS client in " + mode + " mode")
|
||||
import tts_coqui as coqui
|
||||
from tts_coqui import *
|
||||
|
||||
if "edge-tts" in modules:
|
||||
print("Initializing Edge TTS client")
|
||||
@@ -285,7 +290,7 @@ if "chromadb" in modules:
|
||||
)
|
||||
print(f"ChromaDB is remotely configured at {args.chroma_host}:{chroma_port}")
|
||||
|
||||
chromadb_embedder = SentenceTransformer(embedding_model)
|
||||
chromadb_embedder = SentenceTransformer(embedding_model, device=device_string)
|
||||
chromadb_embed_fn = lambda *args, **kwargs: chromadb_embedder.encode(*args, **kwargs).tolist()
|
||||
|
||||
# Check if the db is connected and running, otherwise tell the user
|
||||
@@ -558,6 +563,59 @@ def api_classify_labels():
|
||||
labels = [x["label"] for x in classification]
|
||||
return jsonify({"labels": labels})
|
||||
|
||||
@app.route("/api/coqui-tts/load", methods=["GET"])
|
||||
@require_module("coqui-tts")
|
||||
def load_model():
|
||||
# Accessing the URL parameters
|
||||
_model = request.args.get('_model')
|
||||
|
||||
_gpu = False if args.coqui_gpu else True
|
||||
print(_gpu)
|
||||
|
||||
_progress = request.args.get('_progress')
|
||||
return coqui.load_model(_model, _gpu, _progress)
|
||||
|
||||
@app.route("/api/coqui-tts/list", methods=["GET"]) #dropdown list
|
||||
@require_module("coqui-tts")
|
||||
def coqui_list():
|
||||
return coqui.get_coqui_models()
|
||||
|
||||
@app.route("/api/coqui-tts/multspeaker", methods=["GET"])
|
||||
@require_module("coqui-tts")
|
||||
def is_multi_speaker_model():
|
||||
return coqui.is_multi_speaker_model()
|
||||
|
||||
@app.route("/api/coqui-tts/multlang", methods=["GET"])
|
||||
@require_module("coqui-tts")
|
||||
def is_multi_lang_model():
|
||||
return coqui.is_multi_lang_model()
|
||||
|
||||
@app.route("/api/coqui-tts/speaker_id", methods=["GET"]) #available voices
|
||||
@require_module("coqui-tts")
|
||||
def coqui_download_models():
|
||||
return coqui.get_coqui_download_models()
|
||||
|
||||
@app.route("/api/coqui-tts/checkmap", methods=["GET"]) #checkmap
|
||||
@require_module("coqui-tts")
|
||||
def coqui_checkmap():
|
||||
return coqui.coqui_checkmap()
|
||||
|
||||
@app.route("/api/coqui-tts/download", methods=["GET"])
|
||||
@require_module("coqui-tts")
|
||||
def coqui_modeldownload():
|
||||
_modeldownload = request.args.get('model')
|
||||
return coqui.coqui_modeldownload(_modeldownload)
|
||||
|
||||
@app.route("/api/coqui-tts/tts", methods=["GET"])
|
||||
@require_module("coqui-tts")
|
||||
def coqui_tts():
|
||||
# Accessing the URL parameters
|
||||
text = request.args.get('text')
|
||||
speaker_id = request.args.get('speaker_id')
|
||||
mspker_id = request.args.get('mspker')
|
||||
language_id = request.args.get('language_id')
|
||||
style_wav = request.args.get('style_wav')
|
||||
return coqui.coqui_tts(text, speaker_id, mspker_id, style_wav, language_id)
|
||||
|
||||
@app.route("/api/image", methods=["POST"])
|
||||
@require_module("sd")
|
||||
|
||||
388
tts_coqui.py
Normal file
388
tts_coqui.py
Normal file
@@ -0,0 +1,388 @@
|
||||
import io
|
||||
import asyncio
|
||||
import json
|
||||
import os
|
||||
import torch
|
||||
import gc
|
||||
from pathlib import Path
|
||||
import TTS
|
||||
from TTS.api import TTS
|
||||
from TTS.utils.manage import ModelManager
|
||||
|
||||
from TTS.tts.configs.bark_config import BarkConfig
|
||||
from TTS.tts.models.bark import Bark
|
||||
|
||||
from TTS.tts.configs.tortoise_config import TortoiseConfig
|
||||
from TTS.tts.models.tortoise import Tortoise
|
||||
|
||||
from flask import send_file
|
||||
|
||||
tts = None
|
||||
type = None
|
||||
multlang = "None"
|
||||
multspeak = "None"
|
||||
loadedModel = "None"
|
||||
spkdirectory = ""
|
||||
multspeakjson = ""
|
||||
|
||||
def model_type(_config_path):
|
||||
try:
|
||||
with open(_config_path, 'r') as config_file:
|
||||
config_data = json.load(config_file)
|
||||
|
||||
# Search for the key "model" and print its value
|
||||
if "model" in config_data:
|
||||
model_value = config_data["model"]
|
||||
return model_value
|
||||
else:
|
||||
print("ERR: The key 'model' is not present in the config file.")
|
||||
except FileNotFoundError:
|
||||
print("Config file not found.")
|
||||
except json.JSONDecodeError:
|
||||
pass
|
||||
#print("Invalid JSON format in the config file.")
|
||||
except Exception as e:
|
||||
pass
|
||||
#print("An error occurred:", str(e))
|
||||
|
||||
def load_model(_model, _gpu, _progress):
|
||||
global tts
|
||||
global type
|
||||
global loadedModel
|
||||
global multlang
|
||||
global multspeak
|
||||
|
||||
status = None
|
||||
|
||||
_model_directory, _file = os.path.split(_model)
|
||||
|
||||
if _model_directory == "": #make it assign vars correctly if no filename provioded
|
||||
_model_directory = _file
|
||||
_file = None
|
||||
|
||||
if _model is None:
|
||||
status = "ERROR: Invalid model name or path."
|
||||
else:
|
||||
try:
|
||||
if _gpu == True: #Reclaim memory
|
||||
del tts
|
||||
try:
|
||||
import gc
|
||||
gc.collect()
|
||||
torch.cuda.empty_cache()
|
||||
except Exception:
|
||||
pass
|
||||
except Exception as e:
|
||||
status = str(e)
|
||||
|
||||
_target_directory = ModelManager().output_prefix # models location
|
||||
_modified_speaker_id = _model_directory.replace("\\", "--")
|
||||
|
||||
if _file != None:
|
||||
_model_path = os.path.join(_target_directory, _modified_speaker_id, _file)
|
||||
else:
|
||||
_model_path = os.path.join(_target_directory, _modified_speaker_id)
|
||||
|
||||
_config_path = os.path.join(_target_directory, _modified_speaker_id, "config.json")
|
||||
|
||||
if model_type(_config_path) == "tortoise":
|
||||
print("Loading Tortoise...")
|
||||
print("_model", _model)
|
||||
print("Tortoise not supported at the moment im tired of working on this")
|
||||
#_loadtortoisemodel = _model_directory.replace("--", "/")
|
||||
#print("_loadtortoisemodel", _loadtortoisemodel)
|
||||
|
||||
#config = TortoiseConfig()
|
||||
#model = Tortoise.init_from_config(config)
|
||||
#model.load_checkpoint(config, checkpoint_dir="C:/Users/jsviv/AppData/Local/tts/tts_models--en--multi-dataset--tortoise-v2", eval=False)
|
||||
|
||||
#tts = TTS(_loadtortoisemodel)
|
||||
#tts = TTS(model_name="tts_models/en/multi-dataset/tortoise-v2", progress_bar=True, gpu=True)
|
||||
|
||||
#loadedModel = _model
|
||||
#print("loaded model", loadedModel)
|
||||
|
||||
if model_type(_config_path) == "bark":
|
||||
print("Loading Bark...")
|
||||
_loadbarkmodel = _model_directory.replace("--", "/")
|
||||
tts = TTS(_loadbarkmodel, gpu=_gpu)
|
||||
loadedModel = _model
|
||||
|
||||
_loadertypes = ["tortoise", "bark"]
|
||||
if model_type(_config_path) not in _loadertypes:
|
||||
try:
|
||||
print("Loading ", model_type(_config_path))
|
||||
tts = TTS(model_path=_model_path, config_path=_config_path, progress_bar=_progress, gpu=_gpu)
|
||||
status = "Loaded"
|
||||
loadedModel = _model
|
||||
except Exception as e:
|
||||
print("An exception occurred while loading VITS:", str(e))
|
||||
print("Continuing with other parts of the code...")
|
||||
else:
|
||||
pass
|
||||
|
||||
type = model_type(_config_path)
|
||||
print("Type: ", type)
|
||||
|
||||
if status is None:
|
||||
status = "Unknown error occurred"
|
||||
if type is None:
|
||||
type = "Unknown"
|
||||
|
||||
return status
|
||||
|
||||
def is_multi_speaker_model():
|
||||
global multspeak
|
||||
global type
|
||||
global spkdirectory
|
||||
global multspeakjson
|
||||
|
||||
if tts is None:
|
||||
multspeak = "None"
|
||||
return multspeak
|
||||
try:
|
||||
|
||||
|
||||
if type == "bark":
|
||||
_target_directory = ModelManager().output_prefix
|
||||
# Convert _target_directory to a string and remove the trailing backslash if present
|
||||
_target_directory_str = str(_target_directory)
|
||||
if _target_directory_str.endswith("\\"):
|
||||
_target_directory_str = _target_directory_str[:-1]
|
||||
|
||||
spkdirectory = os.path.join(_target_directory_str, "bark_v0", "speakers")
|
||||
|
||||
subfolder_names = [folder for folder in os.listdir(spkdirectory) if os.path.isdir(os.path.join(spkdirectory, folder))]
|
||||
|
||||
subfolder_names.insert(0, "random") # Add "Random" as the first element in the subfolder_names list
|
||||
|
||||
unique_names = list(dict.fromkeys(subfolder_names))
|
||||
multspeak = json.dumps({index: name for index, name in enumerate(unique_names)})
|
||||
#print(multspeak)
|
||||
else:
|
||||
|
||||
value = tts.speakers
|
||||
if value is not None:
|
||||
unique_speakers = list(dict.fromkeys(value))
|
||||
speaker_dict = {index: value for index, value in enumerate(unique_speakers)}
|
||||
multspeak = json.dumps(speaker_dict)
|
||||
#print(multspeak)
|
||||
else:
|
||||
multspeak = "None"
|
||||
|
||||
|
||||
except Exception as e:
|
||||
print("Error:", e)
|
||||
multspeak = "None"
|
||||
multspeakjson = multspeak
|
||||
return multspeak #return name and ID in named json
|
||||
|
||||
def is_multi_lang_model():
|
||||
global multlang
|
||||
if tts is None:
|
||||
multlang = "None"
|
||||
return multlang
|
||||
try:
|
||||
value = tts.languages
|
||||
if value is not None:
|
||||
unique_lang = list(dict.fromkeys(value))# Remove duplicate values and preserve the order
|
||||
lang_dict = {index: value for index, value in enumerate(unique_lang)} # Create a dictionary with indices as keys and values as keys
|
||||
multlang = json.dumps(lang_dict) # Convert the dictionary to JSON format
|
||||
#print(multlang)
|
||||
else:
|
||||
multlang = "None"
|
||||
except Exception as e:
|
||||
print("Error:", e)
|
||||
multlang = "None"
|
||||
|
||||
return multlang
|
||||
|
||||
def get_coqui_models(): #DROPDOWN MODELS
|
||||
manager = ModelManager()
|
||||
model_folder = manager.output_prefix
|
||||
|
||||
cwd = os.path.dirname(os.path.realpath(__file__))
|
||||
target_directory = model_folder
|
||||
|
||||
if not os.path.exists(target_directory):
|
||||
os.makedirs(target_directory)
|
||||
|
||||
os.chdir(target_directory)
|
||||
folder_list = [
|
||||
folder for folder in os.listdir(target_directory) if os.path.isdir(os.path.join(target_directory, folder)) and "--" in folder and "vocoder" not in folder.lower() and "voice_conversion_models" not in folder.lower()
|
||||
]
|
||||
|
||||
|
||||
file_paths = []
|
||||
|
||||
for folder in folder_list:
|
||||
_config_path = os.path.join(target_directory, folder, "config.json")
|
||||
if model_type(_config_path) == "bark" or model_type(_config_path) == "tortoise":
|
||||
file_paths.append(str(Path(folder, '')))
|
||||
else:
|
||||
for file in os.listdir(os.path.join(target_directory, folder)):
|
||||
if file.endswith(('.pt', '.tar', '.pkl', '.pth')) and not file.startswith('.'):
|
||||
file_paths.append(str(Path(folder, file)))
|
||||
|
||||
merged_json = json.dumps(file_paths)
|
||||
return merged_json
|
||||
|
||||
def coqui_checkmap():
|
||||
manager = ModelManager()
|
||||
model_folder = manager.output_prefix
|
||||
|
||||
cwd = os.path.dirname(os.path.realpath(__file__))
|
||||
target_directory = model_folder
|
||||
|
||||
if not os.path.exists(target_directory):
|
||||
os.makedirs(target_directory)
|
||||
|
||||
os.chdir(target_directory)
|
||||
folder_list = [
|
||||
folder for folder in os.listdir() if os.path.isdir(os.path.join(target_directory, folder)) and "--" in folder and "vocoder" not in folder.lower()
|
||||
]
|
||||
|
||||
file_paths = []
|
||||
|
||||
for folder in folder_list:
|
||||
_config_path = os.path.join(target_directory, folder, "config.json")
|
||||
if model_type(_config_path) == "bark" or model_type(_config_path) == "tortoise":
|
||||
file_paths.append(str(Path(folder, '')))
|
||||
else:
|
||||
for file in os.listdir(os.path.join(target_directory, folder)):
|
||||
if file.endswith(('.pt', '.tar', '.pkl', '.pth')) and not file.startswith('.'):
|
||||
file_paths.append(str(Path(folder, file)))
|
||||
|
||||
# Convert the list into a list of dictionaries with "id" as the key
|
||||
keyed_json_list = [{"id": item} for item in file_paths]
|
||||
|
||||
# Convert the list to a JSON string with indentation
|
||||
keyed_json_string = json.dumps(keyed_json_list, indent=2)
|
||||
|
||||
# Replace double backslashes with single backslashes
|
||||
#keyed_json_string = keyed_json_string.replace("\\\\", "\\")
|
||||
|
||||
return keyed_json_string
|
||||
|
||||
def get_coqui_download_models(): #Avail voices list
|
||||
formatted_list = []
|
||||
#voices_list = json.loads(get_coqui_downloaded())
|
||||
voices_list = TTS.list_models()
|
||||
|
||||
for model in voices_list:
|
||||
split_model = model.split('/')
|
||||
formatted_list.append({
|
||||
"type": split_model[0], #type
|
||||
"lang": split_model[1], #lang
|
||||
"id-only": split_model[2], #id
|
||||
"name-only": split_model[3], #name
|
||||
"id": split_model[0] + '/' + split_model[1] + "/" + split_model[2] + "/" + split_model[3], #combined id and name tts_models/bn/custom/vits-male
|
||||
})
|
||||
|
||||
json_data = json.dumps(formatted_list, indent=4)
|
||||
return json_data
|
||||
|
||||
def coqui_modeldownload(_modeldownload): #Avail voices function
|
||||
print(_modeldownload)
|
||||
try:
|
||||
tts = TTS(model_name=_modeldownload, progress_bar=True, gpu=False)
|
||||
status = "True"
|
||||
except:
|
||||
status = "False"
|
||||
return status
|
||||
|
||||
def coqui_tts(text, speaker_id, mspker_id, style_wav, language_id):
|
||||
global type
|
||||
global multlang
|
||||
global multspeak
|
||||
global loadedModel
|
||||
global spkdirectory
|
||||
global multspeakjson
|
||||
|
||||
try:
|
||||
# Splitting the string to get speaker_id and the rest
|
||||
parts = speaker_id.split("[", 1)
|
||||
speaker_id = parts[0]
|
||||
remainder = parts[1].rstrip("]")
|
||||
variables = remainder.split("][")
|
||||
# Converting to integers with default values of 0 if conversion fails
|
||||
mspker_id = int(variables[0]) if variables[0].isdigit() else 0
|
||||
language_id = int(variables[1]) if variables[1].isdigit() else 0
|
||||
# multspeak = mspker_id # might break previews
|
||||
multlang = language_id
|
||||
except Exception:
|
||||
pass
|
||||
#print("exception 1")
|
||||
|
||||
print("mspker_id: ", mspker_id)
|
||||
print("language_id: ", language_id)
|
||||
|
||||
|
||||
|
||||
try: #see is values passed in URL
|
||||
if language_id is not None:
|
||||
float(language_id)
|
||||
multlang = float(language_id)
|
||||
else:
|
||||
pass
|
||||
except ValueError:
|
||||
pass
|
||||
|
||||
|
||||
try:
|
||||
if mspker_id is not None:
|
||||
float(mspker_id)
|
||||
multspeak = float(mspker_id)
|
||||
else:
|
||||
pass
|
||||
except ValueError:
|
||||
pass
|
||||
|
||||
|
||||
if loadedModel != speaker_id:
|
||||
print("MODEL NOT LOADED!!! Loading... ", loadedModel, speaker_id)
|
||||
load_model(speaker_id, True, True) #use GPU and progress bar?
|
||||
|
||||
|
||||
audio_buffer = io.BytesIO()
|
||||
|
||||
if not isinstance(multspeak, (int, float)) and not isinstance(multlang, (int, float)): #if not a number
|
||||
print("Single Model")
|
||||
tts.tts_to_file(text, file_path=audio_buffer)
|
||||
elif isinstance(multspeak, (int, float)) and not isinstance(multlang, (int, float)):
|
||||
print("speaker only")
|
||||
if type == "bark" or type == "tortoise":
|
||||
try:
|
||||
if multspeakjson == "": #failing because multispeakjson not loaded
|
||||
parsed_multspeak = json.loads(is_multi_speaker_model())
|
||||
else:
|
||||
parsed_multspeak = json.loads(multspeakjson)
|
||||
|
||||
value_at_key = parsed_multspeak.get(str(mspker_id))
|
||||
#print(value_at_key)
|
||||
# ♪ In the jungle, the mighty jungle, the lion barks tonight ♪
|
||||
#I have a silky smooth voice, and today I will tell you about the exercise regimen of the common sloth.
|
||||
if value_at_key == "random":
|
||||
tts.tts_to_file(text, file_path=audio_buffer)
|
||||
else:
|
||||
print("using speaker ", value_at_key)
|
||||
tts.tts_to_file(text, file_path=audio_buffer, voice_dir=spkdirectory, speaker=value_at_key)
|
||||
except Exception as e:
|
||||
print("An error occurred:", str(e))
|
||||
else:
|
||||
tts.tts_to_file(text, speaker=tts.speakers[int(mspker_id)], file_path=audio_buffer)
|
||||
elif not isinstance(multspeak, (int, float)) and isinstance(multlang, (int, float)):
|
||||
print("lang only")
|
||||
tts.tts_to_file(text, language=tts.languages[int(language_id)], file_path=audio_buffer)
|
||||
else:
|
||||
print("spk and lang")
|
||||
tts.tts_to_file(text, speaker=tts.speakers[int(mspker_id)], language=tts.languages[int(language_id)], file_path=audio_buffer)
|
||||
|
||||
audio_buffer.seek(0)
|
||||
response = send_file(audio_buffer, mimetype="audio/wav")
|
||||
|
||||
#reset for next dynamic tts
|
||||
multlang = None
|
||||
multspeak = None
|
||||
return response
|
||||
|
||||
Reference in New Issue
Block a user