Update generic_utils.py (#3561)

Handles cases when git branch produces no output or invalid output. Right now, it just crashes with `StopIteration`

.github/workflows/api_tests.yml

@@ -1,53 +0,0 @@
-name: api_tests
-
-on:
-  push:
-    branches:
-      - main
-jobs:
-  check_skip:
-    runs-on: ubuntu-latest
-    if: "! contains(github.event.head_commit.message, '[ci skip]')"
-    steps:
-      - run: echo "${{ github.event.head_commit.message }}"
-
-  test:
-    runs-on: ubuntu-latest
-    strategy:
-      fail-fast: false
-      matrix:
-        python-version: [3.9, "3.10", "3.11"]
-        experimental: [false]
-    steps:
-      - uses: actions/checkout@v3
-      - name: Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@v4
-        with:
-          python-version: ${{ matrix.python-version }}
-          architecture: x64
-          cache: 'pip'
-          cache-dependency-path: 'requirements*'
-      - name: check OS
-        run: cat /etc/os-release
-      - name: set ENV
-        run: |
-          export TRAINER_TELEMETRY=0
-      - name: Install dependencies
-        run: |
-          sudo apt-get update
-          sudo apt-get install -y --no-install-recommends git make gcc
-          sudo apt-get install espeak-ng
-          make system-deps
-      - name: Install/upgrade Python setup deps
-        run: python3 -m pip install --upgrade pip setuptools wheel
-      - name: Replace scarf urls
-        run: |
-          sed -i 's/https:\/\/coqui.gateway.scarf.sh\//https:\/\/github.com\/coqui-ai\/TTS\/releases\/download\//g' TTS/.models.json
-      - name: Install TTS
-        run: |
-          python3 -m pip install .[all]
-          python3 setup.py egg_info
-      - name: Unit tests
-        run: make api_tests
-        env:
-          COQUI_STUDIO_TOKEN: ${{ secrets.COQUI_STUDIO_TOKEN }}

.github/workflows/zoo_tests_tortoise.yml

@@ -1,52 +0,0 @@
-name: zoo-tests-tortoise
-
-on:
-  push:
-    branches:
-      - main
-  pull_request:
-    types: [opened, synchronize, reopened]
-jobs:
-  check_skip:
-    runs-on: ubuntu-latest
-    if: "! contains(github.event.head_commit.message, '[ci skip]')"
-    steps:
-      - run: echo "${{ github.event.head_commit.message }}"
-
-  test:
-    runs-on: ubuntu-latest
-    strategy:
-      fail-fast: false
-      matrix:
-        python-version: [3.9, "3.10", "3.11"]
-        experimental: [false]
-    steps:
-      - uses: actions/checkout@v3
-      - name: Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@v4
-        with:
-          python-version: ${{ matrix.python-version }}
-          architecture: x64
-          cache: 'pip'
-          cache-dependency-path: 'requirements*'
-      - name: check OS
-        run: cat /etc/os-release
-      - name: set ENV
-        run: export TRAINER_TELEMETRY=0
-      - name: Install dependencies
-        run: |
-          sudo apt-get update
-          sudo apt-get install -y git make gcc
-          sudo apt-get install espeak espeak-ng
-          make system-deps
-      - name: Install/upgrade Python setup deps
-        run: python3 -m pip install --upgrade pip setuptools wheel
-      - name: Replace scarf urls
-        run: |
-          sed -i 's/https:\/\/coqui.gateway.scarf.sh\//https:\/\/github.com\/coqui-ai\/TTS\/releases\/download\//g' TTS/.models.json
-      - name: Install TTS
-        run: |
-          python3 -m pip install .[all]
-          python3 setup.py egg_info
-      - name: Unit tests
-        run: nose2 -F -v -B --with-coverage --coverage TTS tests.zoo_tests.test_models.test_tortoise

CONTRIBUTING.md

@@ -48,7 +48,7 @@ The following steps are tested on an Ubuntu system.
 
 1. Fork 🐸TTS[https://github.com/coqui-ai/TTS] by clicking the fork button at the top right corner of the project page.
 
-2. Clone 🐸TTS and add the main repo as a new remote named ```upsteam```.
+2. Clone 🐸TTS and add the main repo as a new remote named ```upstream```.
 
     ```bash
     $ git clone git@github.com:<your Github name>/TTS.git

Makefile

@@ -35,9 +35,6 @@ test_zoo:	## run zoo tests.
 inference_tests: ## run inference tests.
 	nose2 -F -v -B --with-coverage --coverage TTS tests.inference_tests
 
-api_tests: ## run api tests.
-	nose2 -F -v -B --with-coverage --coverage TTS tests.api_tests
-
 data_tests: ## run data tests.
 	nose2 -F -v -B --with-coverage --coverage TTS tests.data_tests
 

README.md

@@ -7,11 +7,6 @@
 - 📣 [🐶Bark](https://github.com/suno-ai/bark) is now available for inference with unconstrained voice cloning. [Docs](https://tts.readthedocs.io/en/dev/models/bark.html)
 - 📣 You can use [~1100 Fairseq models](https://github.com/facebookresearch/fairseq/tree/main/examples/mms) with 🐸TTS.
 - 📣 🐸TTS now supports 🐢Tortoise with faster inference. [Docs](https://tts.readthedocs.io/en/dev/models/tortoise.html)
-- 📣 **Coqui Studio API** is landed on 🐸TTS. - [Example](https://github.com/coqui-ai/TTS/blob/dev/README.md#-python-api)
-- 📣 [**Coqui Studio API**](https://docs.coqui.ai/docs) is live.
-- 📣 Voice generation with prompts - **Prompt to Voice** - is live on [**Coqui Studio**](https://app.coqui.ai/auth/signin)!! - [Blog Post](https://coqui.ai/blog/tts/prompt-to-voice)
-- 📣 Voice generation with fusion - **Voice fusion** - is live on [**Coqui Studio**](https://app.coqui.ai/auth/signin).
-- 📣 Voice cloning is live on [**Coqui Studio**](https://app.coqui.ai/auth/signin).
 
 <div align="center">
 <img src="https://static.scarf.sh/a.png?x-pxid=cf317fe7-2188-4721-bc01-124bb5d5dbb2" />
@@ -253,29 +248,6 @@ tts.tts_with_vc_to_file(
 )
 ```
 
-#### Example using [🐸Coqui Studio](https://coqui.ai) voices.
-You access all of your cloned voices and built-in speakers in [🐸Coqui Studio](https://coqui.ai).
-To do this, you'll need an API token, which you can obtain from the [account page](https://coqui.ai/account).
-After obtaining the API token, you'll need to configure the COQUI_STUDIO_TOKEN environment variable.
-
-Once you have a valid API token in place, the studio speakers will be displayed as distinct models within the list.
-These models will follow the naming convention `coqui_studio/en/<studio_speaker_name>/coqui_studio`
-
-```python
-# XTTS model
-models = TTS(cs_api_model="XTTS").list_models()
-# Init TTS with the target studio speaker
-tts = TTS(model_name="coqui_studio/en/Torcull Diarmuid/coqui_studio", progress_bar=False)
-# Run TTS
-tts.tts_to_file(text="This is a test.", language="en", file_path=OUTPUT_PATH)
-
-# V1 model
-models = TTS(cs_api_model="V1").list_models()
-# Run TTS with emotion and speed control
-# Emotion control only works with V1 model
-tts.tts_to_file(text="This is a test.", file_path=OUTPUT_PATH, emotion="Happy", speed=1.5)
-```
-
 #### Example text to speech using **Fairseq models in ~1100 languages** 🤯.
 For Fairseq models, use the following name format: `tts_models/<lang-iso_code>/fairseq/vits`.
 You can find the language ISO codes [here](https://dl.fbaipublicfiles.com/mms/tts/all-tts-languages.html)
@@ -351,12 +323,6 @@ If you don't specify any models, then it uses LJSpeech based English model.
   $ tts --text "Text for TTS" --pipe_out --out_path output/path/speech.wav | aplay
   ```
 
-- Run TTS and define speed factor to use for 🐸Coqui Studio models, between 0.0 and 2.0:
-
-  ```
-  $ tts --text "Text for TTS" --model_name "coqui_studio/<language>/<dataset>/<model_name>" --speed 1.2 --out_path output/path/speech.wav
-  ```
-
 - Run a TTS model with its default vocoder model:
 
   ```

TTS/.models.json

@@ -3,12 +3,13 @@
         "multilingual": {
             "multi-dataset": {
                 "xtts_v2": {
-                    "description": "XTTS-v2.0.2 by Coqui with 16 languages.",
+                    "description": "XTTS-v2.0.3 by Coqui with 17 languages.",
                     "hf_url": [
                         "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/model.pth",
                         "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/config.json",
                         "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/vocab.json",
-                        "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/hash.md5"
+                        "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/hash.md5",
+                        "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/speakers_xtts.pth"
                     ],
                     "model_hash": "10f92b55c512af7a8d39d650547a15a7",
                     "default_vocoder": null,
@@ -45,7 +46,7 @@
                     "hf_url": [
                         "https://coqui.gateway.scarf.sh/hf/bark/coarse_2.pt",
                         "https://coqui.gateway.scarf.sh/hf/bark/fine_2.pt",
-                        "https://app.coqui.ai/tts_model/text_2.pt",
+                        "https://coqui.gateway.scarf.sh/hf/text_2.pt",
                         "https://coqui.gateway.scarf.sh/hf/bark/config.json",
                         "https://coqui.gateway.scarf.sh/hf/bark/hubert.pt",
                         "https://coqui.gateway.scarf.sh/hf/bark/tokenizer.pth"
@@ -270,7 +271,7 @@
                 "tortoise-v2": {
                     "description": "Tortoise tts model https://github.com/neonbjb/tortoise-tts",
                     "github_rls_url": [
-                        "https://app.coqui.ai/tts_model/autoregressive.pth",
+                        "https://coqui.gateway.scarf.sh/v0.14.1_models/autoregressive.pth",
                         "https://coqui.gateway.scarf.sh/v0.14.1_models/clvp2.pth",
                         "https://coqui.gateway.scarf.sh/v0.14.1_models/cvvp.pth",
                         "https://coqui.gateway.scarf.sh/v0.14.1_models/diffusion_decoder.pth",

TTS/VERSION

@@ -1 +1 @@
-0.21.3
+0.22.0

TTS/api.py

@@ -6,7 +6,6 @@ from typing import Union
 import numpy as np
 from torch import nn
 
-from TTS.cs_api import CS_API
 from TTS.utils.audio.numpy_transforms import save_wav
 from TTS.utils.manage import ModelManager
 from TTS.utils.synthesizer import Synthesizer
@@ -24,7 +23,6 @@ class TTS(nn.Module):
         vocoder_path: str = None,
         vocoder_config_path: str = None,
         progress_bar: bool = True,
-        cs_api_model: str = "XTTS",
         gpu=False,
     ):
         """🐸TTS python interface that allows to load and use the released models.
@@ -60,9 +58,6 @@ class TTS(nn.Module):
             vocoder_path (str, optional): Path to the vocoder checkpoint. Defaults to None.
             vocoder_config_path (str, optional): Path to the vocoder config. Defaults to None.
             progress_bar (bool, optional): Whether to pring a progress bar while downloading a model. Defaults to True.
-            cs_api_model (str, optional): Name of the model to use for the Coqui Studio API. Available models are
-                "XTTS", "V1". You can also use `TTS.cs_api.CS_API" for more control.
-                Defaults to "XTTS".
             gpu (bool, optional): Enable/disable GPU. Some models might be too slow on CPU. Defaults to False.
         """
         super().__init__()
@@ -70,14 +65,12 @@ class TTS(nn.Module):
         self.config = load_config(config_path) if config_path else None
         self.synthesizer = None
         self.voice_converter = None
-        self.csapi = None
-        self.cs_api_model = cs_api_model
         self.model_name = ""
         if gpu:
             warnings.warn("`gpu` will be deprecated. Please use `tts.to(device)` instead.")
 
         if model_name is not None and len(model_name) > 0:
-            if "tts_models" in model_name or "coqui_studio" in model_name:
+            if "tts_models" in model_name:
                 self.load_tts_model_by_name(model_name, gpu)
             elif "voice_conversion_models" in model_name:
                 self.load_vc_model_by_name(model_name, gpu)
@@ -99,12 +92,6 @@ class TTS(nn.Module):
             return self.synthesizer.tts_model.speaker_manager.num_speakers > 1
         return False
 
-    @property
-    def is_coqui_studio(self):
-        if self.model_name is None:
-            return False
-        return "coqui_studio" in self.model_name
-
     @property
     def is_multi_lingual(self):
         # Not sure what sets this to None, but applied a fix to prevent crashing.
@@ -136,14 +123,7 @@ class TTS(nn.Module):
         return Path(__file__).parent / ".models.json"
 
     def list_models(self):
-        try:
-            csapi = CS_API(model=self.cs_api_model)
-            models = csapi.list_speakers_as_tts_models()
-        except ValueError as e:
-            print(e)
-            models = []
-        manager = ModelManager(models_file=TTS.get_models_file_path(), progress_bar=False, verbose=False)
-        return manager.list_tts_models() + models
+        return ModelManager(models_file=TTS.get_models_file_path(), progress_bar=False, verbose=False)
 
     def download_model_by_name(self, model_name: str):
         model_path, config_path, model_item = self.manager.download_model(model_name)
@@ -186,12 +166,8 @@ class TTS(nn.Module):
         TODO: Add tests
         """
         self.synthesizer = None
-        self.csapi = None
         self.model_name = model_name
 
-        if "coqui_studio" in model_name:
-            self.csapi = CS_API()
-        else:
         model_path, config_path, vocoder_path, vocoder_config_path, model_dir = self.download_model_by_name(
             model_name
         )
@@ -246,7 +222,6 @@ class TTS(nn.Module):
         **kwargs,
     ) -> None:
         """Check if the arguments are valid for the model."""
-        if not self.is_coqui_studio:
         # check for the coqui tts models
         if self.is_multi_speaker and (speaker is None and speaker_wav is None):
             raise ValueError("Model is multi-speaker but no `speaker` is provided.")
@@ -257,66 +232,7 @@ class TTS(nn.Module):
         if not self.is_multi_lingual and language is not None:
             raise ValueError("Model is not multi-lingual but `language` is provided.")
         if not emotion is None and not speed is None:
-                raise ValueError("Emotion and speed can only be used with Coqui Studio models.")
-        else:
-            if emotion is None:
-                emotion = "Neutral"
-            if speed is None:
-                speed = 1.0
-            # check for the studio models
-            if speaker_wav is not None:
-                raise ValueError("Coqui Studio models do not support `speaker_wav` argument.")
-            if speaker is not None:
-                raise ValueError("Coqui Studio models do not support `speaker` argument.")
-            if language is not None and language != "en":
-                raise ValueError("Coqui Studio models currently support only `language=en` argument.")
-            if emotion not in ["Neutral", "Happy", "Sad", "Angry", "Dull"]:
-                raise ValueError(f"Emotion - `{emotion}` - must be one of `Neutral`, `Happy`, `Sad`, `Angry`, `Dull`.")
-
-    def tts_coqui_studio(
-        self,
-        text: str,
-        speaker_name: str = None,
-        language: str = None,
-        emotion: str = None,
-        speed: float = 1.0,
-        pipe_out=None,
-        file_path: str = None,
-    ) -> Union[np.ndarray, str]:
-        """Convert text to speech using Coqui Studio models. Use `CS_API` class if you are only interested in the API.
-
-        Args:
-            text (str):
-                Input text to synthesize.
-            speaker_name (str, optional):
-                Speaker name from Coqui Studio. Defaults to None.
-            language (str): Language of the text. If None, the default language of the speaker is used. Language is only
-                supported by `XTTS` model.
-            emotion (str, optional):
-                Emotion of the speaker. One of "Neutral", "Happy", "Sad", "Angry", "Dull". Emotions are only available
-                with "V1" model. Defaults to None.
-            speed (float, optional):
-                Speed of the speech. Defaults to 1.0.
-            pipe_out (BytesIO, optional):
-                Flag to stdout the generated TTS wav file for shell pipe.
-            file_path (str, optional):
-                Path to save the output file. When None it returns the `np.ndarray` of waveform. Defaults to None.
-
-        Returns:
-            Union[np.ndarray, str]: Waveform of the synthesized speech or path to the output file.
-        """
-        speaker_name = self.model_name.split("/")[2]
-        if file_path is not None:
-            return self.csapi.tts_to_file(
-                text=text,
-                speaker_name=speaker_name,
-                language=language,
-                speed=speed,
-                pipe_out=pipe_out,
-                emotion=emotion,
-                file_path=file_path,
-            )[0]
-        return self.csapi.tts(text=text, speaker_name=speaker_name, language=language, speed=speed, emotion=emotion)[0]
+            raise ValueError("Emotion and speed can only be used with Coqui Studio models. Which is discontinued.")
 
     def tts(
         self,
@@ -357,10 +273,6 @@ class TTS(nn.Module):
         self._check_arguments(
             speaker=speaker, language=language, speaker_wav=speaker_wav, emotion=emotion, speed=speed, **kwargs
         )
-        if self.csapi is not None:
-            return self.tts_coqui_studio(
-                text=text, speaker_name=speaker, language=language, emotion=emotion, speed=speed
-            )
         wav = self.synthesizer.tts(
             text=text,
             speaker_name=speaker,
@@ -419,16 +331,6 @@ class TTS(nn.Module):
         """
         self._check_arguments(speaker=speaker, language=language, speaker_wav=speaker_wav, **kwargs)
 
-        if self.csapi is not None:
-            return self.tts_coqui_studio(
-                text=text,
-                speaker_name=speaker,
-                language=language,
-                emotion=emotion,
-                speed=speed,
-                file_path=file_path,
-                pipe_out=pipe_out,
-            )
         wav = self.tts(
             text=text,
             speaker=speaker,

TTS/bin/synthesize.py

@@ -66,12 +66,6 @@ If you don't specify any models, then it uses LJSpeech based English model.
   $ tts --text "Text for TTS" --pipe_out --out_path output/path/speech.wav | aplay
   ```
 
-- Run TTS and define speed factor to use for 🐸Coqui Studio models, between 0.0 and 2.0:
-
-  ```
-  $ tts --text "Text for TTS" --model_name "coqui_studio/<language>/<dataset>/<model_name>" --speed 1.2 --out_path output/path/speech.wav
-  ```
-
 - Run a TTS model with its default vocoder model:
 
   ```
@@ -222,25 +216,6 @@ def main():
         default=None,
     )
     parser.add_argument("--encoder_config_path", type=str, help="Path to speaker encoder config file.", default=None)
-
-    # args for coqui studio
-    parser.add_argument(
-        "--cs_model",
-        type=str,
-        help="Name of the 🐸Coqui Studio model. Available models are `XTTS`, `V1`.",
-    )
-    parser.add_argument(
-        "--emotion",
-        type=str,
-        help="Emotion to condition the model with. Only available for 🐸Coqui Studio `V1` model.",
-        default=None,
-    )
-    parser.add_argument(
-        "--language",
-        type=str,
-        help="Language to condition the model with. Only available for 🐸Coqui Studio `XTTS` model.",
-        default=None,
-    )
     parser.add_argument(
         "--pipe_out",
         help="stdout the generated TTS wav file for shell pipe.",
@@ -249,12 +224,6 @@ def main():
         const=True,
         default=False,
     )
-    parser.add_argument(
-        "--speed",
-        type=float,
-        help="Speed factor to use for 🐸Coqui Studio models, between 0.0 and 2.0.",
-        default=None,
-    )
     
     # args for multi-speaker synthesis
     parser.add_argument("--speakers_file_path", type=str, help="JSON file for multi-speaker model.", default=None)
@@ -389,7 +358,6 @@ def main():
 
         # CASE1 #list : list pre-trained TTS models
         if args.list_models:
-            manager.add_cs_api_models(api.list_models())
             manager.list_models()
             sys.exit()
 
@@ -404,29 +372,7 @@ def main():
             manager.model_info_by_full_name(model_query_full_name)
             sys.exit()
 
-        # CASE3: TTS with coqui studio models
-        if "coqui_studio" in args.model_name:
-            print(" > Using 🐸Coqui Studio model: ", args.model_name)
-            api = TTS(model_name=args.model_name, cs_api_model=args.cs_model)
-            api.tts_to_file(
-                text=args.text,
-                emotion=args.emotion,
-                file_path=args.out_path,
-                language=args.language,
-                speed=args.speed,
-                pipe_out=pipe_out,
-            )
-            print(" > Saving output to ", args.out_path)
-            return
-
-        if args.language_idx is None and args.language is not None:
-            msg = (
-                "--language is only supported for Coqui Studio models. "
-                "Use --language_idx to specify the target language for multilingual models."
-            )
-            raise ValueError(msg)
-
-        # CASE4: load pre-trained model paths
+        # CASE3: load pre-trained model paths
         if args.model_name is not None and not args.model_path:
             model_path, config_path, model_item = manager.download_model(args.model_name)
             # tts model
@@ -454,7 +400,7 @@ def main():
         if args.vocoder_name is not None and not args.vocoder_path:
             vocoder_path, vocoder_config_path, _ = manager.download_model(args.vocoder_name)
 
-        # CASE5: set custom model paths
+        # CASE4: set custom model paths
         if args.model_path is not None:
             tts_path = args.model_path
             tts_config_path = args.config_path

TTS/bin/train_encoder.py

@@ -125,7 +125,7 @@ def evaluation(model, criterion, data_loader, global_step):
 
 def train(model, optimizer, scheduler, criterion, data_loader, eval_data_loader, global_step):
     model.train()
-    best_loss = float("inf")
+    best_loss = {"train_loss": None, "eval_loss": float("inf")}
     avg_loader_time = 0
     end_time = time.time()
     for epoch in range(c.epochs):
@@ -248,7 +248,7 @@ def train(model, optimizer, scheduler, criterion, data_loader, eval_data_loader,
             )
             # save the best checkpoint
             best_loss = save_best_model(
-                eval_loss,
+                {"train_loss": None, "eval_loss": eval_loss},
                 best_loss,
                 c,
                 model,

TTS/config/__init__.py

@@ -16,12 +16,9 @@ def read_json_with_comments(json_path):
     # fallback to json
     with fsspec.open(json_path, "r", encoding="utf-8") as f:
         input_str = f.read()
-    # handle comments
-    input_str = re.sub(r"\\\n", "", input_str)
-    input_str = re.sub(r"//.*\n", "\n", input_str)
-    data = json.loads(input_str)
-    return data
-
+    # handle comments but not urls with //
+    input_str = re.sub(r"(\"(?:[^\"\\]|\\.)*\")|(/\*(?:.|[\\n\\r])*?\*/)|(//.*)", lambda m: m.group(1) or m.group(2) or "", input_str)
+    return json.loads(input_str)
 
 def register_config(model_name: str) -> Coqpit:
     """Find the right config for the given model name.

TTS/cs_api.py

@@ -1,317 +0,0 @@
-import http.client
-import json
-import os
-import tempfile
-import urllib.request
-from typing import Tuple
-
-import numpy as np
-import requests
-from scipy.io import wavfile
-
-from TTS.utils.audio.numpy_transforms import save_wav
-
-
-class Speaker(object):
-    """Convert dict to object."""
-
-    def __init__(self, d, is_voice=False):
-        self.is_voice = is_voice
-        for k, v in d.items():
-            if isinstance(k, (list, tuple)):
-                setattr(self, k, [Speaker(x) if isinstance(x, dict) else x for x in v])
-            else:
-                setattr(self, k, Speaker(v) if isinstance(v, dict) else v)
-
-    def __repr__(self):
-        return str(self.__dict__)
-
-
-class CS_API:
-    """🐸Coqui Studio API Wrapper.
-
-    🐸Coqui Studio is the most advanced voice generation platform. You can generate new voices by voice cloning, voice
-    interpolation, or our unique prompt to voice technology. It also provides a set of built-in voices with different
-    characteristics. You can use these voices to generate new audio files or use them in your applications.
-    You can use all the built-in and your own 🐸Coqui Studio speakers with this API with an API token.
-    You can signup to 🐸Coqui Studio from https://app.coqui.ai/auth/signup and get an API token from
-    https://app.coqui.ai/account. We can either enter the token as an environment variable as
-    `export COQUI_STUDIO_TOKEN=<token>` or pass it as `CS_API(api_token=<toke>)`.
-    Visit https://app.coqui.ai/api for more information.
-
-
-    Args:
-        api_token (str): 🐸Coqui Studio API token. If not provided, it will be read from the environment variable
-            `COQUI_STUDIO_TOKEN`.
-        model (str): 🐸Coqui Studio model. It can be either `V1`, `XTTS`. Default is `XTTS`.
-
-
-    Example listing all available speakers:
-        >>> from TTS.api import CS_API
-        >>> tts = CS_API()
-        >>> tts.speakers
-
-    Example listing all emotions:
-        >>> # emotions are only available for `V1` model
-        >>> from TTS.api import CS_API
-        >>> tts = CS_API(model="V1")
-        >>> tts.emotions
-
-    Example with a built-in 🐸 speaker:
-        >>> from TTS.api import CS_API
-        >>> tts = CS_API()
-        >>> wav, sr = api.tts("Hello world", speaker_name=tts.speakers[0].name)
-        >>> filepath = tts.tts_to_file(text="Hello world!", speaker_name=tts.speakers[0].name, file_path="output.wav")
-
-    Example with multi-language model:
-        >>> from TTS.api import CS_API
-        >>> tts = CS_API(model="XTTS")
-        >>> wav, sr = api.tts("Hello world", speaker_name=tts.speakers[0].name, language="en")
-    """
-
-    MODEL_ENDPOINTS = {
-        "V1": {
-            "list_speakers": "https://app.coqui.ai/api/v2/speakers",
-            "synthesize": "https://app.coqui.ai/api/v2/samples",
-            "list_voices": "https://app.coqui.ai/api/v2/voices",
-        },
-        "XTTS": {
-            "list_speakers": "https://app.coqui.ai/api/v2/speakers",
-            "synthesize": "https://app.coqui.ai/api/v2/samples/xtts/render/",
-            "list_voices": "https://app.coqui.ai/api/v2/voices/xtts",
-        },
-    }
-
-    SUPPORTED_LANGUAGES = ["en", "es", "de", "fr", "it", "pt", "pl", "tr", "ru", "nl", "cs", "ar", "zh-cn", "ja"]
-
-    def __init__(self, api_token=None, model="XTTS"):
-        self.api_token = api_token
-        self.model = model
-        self.headers = None
-        self._speakers = None
-        self._check_token()
-
-    @staticmethod
-    def ping_api():
-        URL = "https://coqui.gateway.scarf.sh/tts/api"
-        _ = requests.get(URL)
-
-    @property
-    def speakers(self):
-        if self._speakers is None:
-            self._speakers = self.list_all_speakers()
-        return self._speakers
-
-    @property
-    def emotions(self):
-        """Return a list of available emotions.
-
-        TODO: Get this from the API endpoint.
-        """
-        if self.model == "V1":
-            return ["Neutral", "Happy", "Sad", "Angry", "Dull"]
-        else:
-            raise ValueError(f"❗ Emotions are not available for {self.model}.")
-
-    def _check_token(self):
-        if self.api_token is None:
-            self.api_token = os.environ.get("COQUI_STUDIO_TOKEN")
-            self.headers = {"Content-Type": "application/json", "Authorization": f"Bearer {self.api_token}"}
-        if not self.api_token:
-            raise ValueError(
-                "No API token found for 🐸Coqui Studio voices - https://coqui.ai \n"
-                "Visit 🔗https://app.coqui.ai/account to get one.\n"
-                "Set it as an environment variable `export COQUI_STUDIO_TOKEN=<token>`\n"
-                ""
-            )
-
-    def list_all_speakers(self):
-        """Return both built-in Coqui Studio speakers and custom voices created by the user."""
-        return self.list_speakers() + self.list_voices()
-
-    def list_speakers(self):
-        """List built-in Coqui Studio speakers."""
-        self._check_token()
-        conn = http.client.HTTPSConnection("app.coqui.ai")
-        url = self.MODEL_ENDPOINTS[self.model]["list_speakers"]
-        conn.request("GET", f"{url}?page=1&per_page=100", headers=self.headers)
-        res = conn.getresponse()
-        data = res.read()
-        return [Speaker(s) for s in json.loads(data)["result"]]
-
-    def list_voices(self):
-        """List custom voices created by the user."""
-        conn = http.client.HTTPSConnection("app.coqui.ai")
-        url = self.MODEL_ENDPOINTS[self.model]["list_voices"]
-        conn.request("GET", f"{url}?page=1&per_page=100", headers=self.headers)
-        res = conn.getresponse()
-        data = res.read()
-        return [Speaker(s, True) for s in json.loads(data)["result"]]
-
-    def list_speakers_as_tts_models(self):
-        """List speakers in ModelManager format."""
-        models = []
-        for speaker in self.speakers:
-            model = f"coqui_studio/multilingual/{speaker.name}/{self.model}"
-            models.append(model)
-        return models
-
-    def name_to_speaker(self, name):
-        for speaker in self.speakers:
-            if speaker.name == name:
-                return speaker
-        raise ValueError(f"Speaker {name} not found in {self.speakers}")
-
-    def id_to_speaker(self, speaker_id):
-        for speaker in self.speakers:
-            if speaker.id == speaker_id:
-                return speaker
-        raise ValueError(f"Speaker {speaker_id} not found.")
-
-    @staticmethod
-    def url_to_np(url):
-        tmp_file, _ = urllib.request.urlretrieve(url)
-        rate, data = wavfile.read(tmp_file)
-        return data, rate
-
-    @staticmethod
-    def _create_payload(model, text, speaker, speed, emotion, language):
-        payload = {}
-        # if speaker.is_voice:
-        payload["voice_id"] = speaker.id
-        # else:
-        payload["speaker_id"] = speaker.id
-
-        if model == "V1":
-            payload.update(
-                {
-                    "emotion": emotion,
-                    "name": speaker.name,
-                    "text": text,
-                    "speed": speed,
-                }
-            )
-        elif model == "XTTS":
-            payload.update(
-                {
-                    "name": speaker.name,
-                    "text": text,
-                    "speed": speed,
-                    "language": language,
-                }
-            )
-        else:
-            raise ValueError(f"❗ Unknown model {model}")
-        return payload
-
-    def _check_tts_args(self, text, speaker_name, speaker_id, emotion, speed, language):
-        assert text is not None, "❗ text is required for V1 model."
-        assert speaker_name is not None, "❗ speaker_name is required for V1 model."
-        if self.model == "V1":
-            if emotion is None:
-                emotion = "Neutral"
-            assert language is None, "❗ language is not supported for V1 model."
-        elif self.model == "XTTS":
-            assert emotion is None, f"❗ Emotions are not supported for XTTS model. Use V1 model."
-            assert language is not None, "❗ Language is required for XTTS model."
-            assert (
-                language in self.SUPPORTED_LANGUAGES
-            ), f"❗ Language {language} is not yet supported. Check https://docs.coqui.ai/reference/samples_xtts_create."
-        return text, speaker_name, speaker_id, emotion, speed, language
-
-    def tts(
-        self,
-        text: str,
-        speaker_name: str = None,
-        speaker_id=None,
-        emotion=None,
-        speed=1.0,
-        language=None,  # pylint: disable=unused-argument
-    ) -> Tuple[np.ndarray, int]:
-        """Synthesize speech from text.
-
-        Args:
-            text (str): Text to synthesize.
-            speaker_name (str): Name of the speaker. You can get the list of speakers with `list_speakers()` and
-                voices (user generated speakers) with `list_voices()`.
-            speaker_id (str): Speaker ID. If None, the speaker name is used.
-            emotion (str): Emotion of the speaker. One of "Neutral", "Happy", "Sad", "Angry", "Dull". Emotions are only
-                supported by `V1` model. Defaults to None.
-            speed (float): Speed of the speech. 1.0 is normal speed.
-            language (str): Language of the text. If None, the default language of the speaker is used. Language is only
-                supported by `XTTS` model. See https://docs.coqui.ai/reference/samples_xtts_create for supported languages.
-        """
-        self._check_token()
-        self.ping_api()
-
-        if speaker_name is None and speaker_id is None:
-            raise ValueError(" [!] Please provide either a `speaker_name` or a `speaker_id`.")
-        if speaker_id is None:
-            speaker = self.name_to_speaker(speaker_name)
-        else:
-            speaker = self.id_to_speaker(speaker_id)
-
-        text, speaker_name, speaker_id, emotion, speed, language = self._check_tts_args(
-            text, speaker_name, speaker_id, emotion, speed, language
-        )
-
-        conn = http.client.HTTPSConnection("app.coqui.ai")
-        payload = self._create_payload(self.model, text, speaker, speed, emotion, language)
-        url = self.MODEL_ENDPOINTS[self.model]["synthesize"]
-        conn.request("POST", url, json.dumps(payload), self.headers)
-        res = conn.getresponse()
-        data = res.read()
-        try:
-            wav, sr = self.url_to_np(json.loads(data)["audio_url"])
-        except KeyError as e:
-            raise ValueError(f" [!] 🐸 API returned error: {data}") from e
-        return wav, sr
-
-    def tts_to_file(
-        self,
-        text: str,
-        speaker_name: str,
-        speaker_id=None,
-        emotion=None,
-        speed=1.0,
-        pipe_out=None,
-        language=None,
-        file_path: str = None,
-    ) -> str:
-        """Synthesize speech from text and save it to a file.
-
-        Args:
-            text (str): Text to synthesize.
-            speaker_name (str): Name of the speaker. You can get the list of speakers with `list_speakers()` and
-                voices (user generated speakers) with `list_voices()`.
-            speaker_id (str): Speaker ID. If None, the speaker name is used.
-            emotion (str): Emotion of the speaker. One of "Neutral", "Happy", "Sad", "Angry", "Dull".
-            speed (float): Speed of the speech. 1.0 is normal speed.
-            pipe_out (BytesIO, optional): Flag to stdout the generated TTS wav file for shell pipe.
-            language (str): Language of the text. If None, the default language of the speaker is used. Language is only
-                supported by `XTTS` model. Currently supports en, de, es, fr, it, pt, pl. Defaults to "en".
-            file_path (str): Path to save the file. If None, a temporary file is created.
-        """
-        if file_path is None:
-            file_path = tempfile.mktemp(".wav")
-        wav, sr = self.tts(text, speaker_name, speaker_id, emotion, speed, language)
-        save_wav(wav=wav, path=file_path, sample_rate=sr, pipe_out=pipe_out)
-        return file_path
-
-
-if __name__ == "__main__":
-    import time
-
-    api = CS_API()
-    print(api.speakers)
-    print(api.list_speakers_as_tts_models())
-
-    ts = time.time()
-    wav, sr = api.tts(
-        "It took me quite a long time to develop a voice.", language="en", speaker_name=api.speakers[0].name
-    )
-    print(f" [i] XTTS took {time.time() - ts:.2f}s")
-
-    filepath = api.tts_to_file(
-        text="Hello world!", speaker_name=api.speakers[0].name, language="en", file_path="output.wav"
-    )

TTS/tts/datasets/dataset.py

@@ -13,6 +13,8 @@ from TTS.tts.utils.data import prepare_data, prepare_stop_target, prepare_tensor
 from TTS.utils.audio import AudioProcessor
 from TTS.utils.audio.numpy_transforms import compute_energy as calculate_energy
 
+import mutagen
+
 # to prevent too many open files error as suggested here
 # https://github.com/pytorch/pytorch/issues/11201#issuecomment-421146936
 torch.multiprocessing.set_sharing_strategy("file_system")
@@ -42,6 +44,15 @@ def string2filename(string):
     return filename
 
 
+def get_audio_size(audiopath):
+    extension = audiopath.rpartition(".")[-1].lower()
+    if extension not in {"mp3", "wav", "flac"}:
+        raise RuntimeError(f"The audio format {extension} is not supported, please convert the audio files to mp3, flac, or wav format!")
+
+    audio_info = mutagen.File(audiopath).info
+    return int(audio_info.length * audio_info.sample_rate)
+
+
 class TTSDataset(Dataset):
     def __init__(
         self,
@@ -176,7 +187,7 @@ class TTSDataset(Dataset):
         lens = []
         for item in self.samples:
             _, wav_file, *_ = _parse_sample(item)
-            audio_len = os.path.getsize(wav_file) / 16 * 8  # assuming 16bit audio
+            audio_len = get_audio_size(wav_file)
             lens.append(audio_len)
         return lens
 
@@ -295,7 +306,7 @@ class TTSDataset(Dataset):
     def _compute_lengths(samples):
         new_samples = []
         for item in samples:
-            audio_length = os.path.getsize(item["audio_file"]) / 16 * 8  # assuming 16bit audio
+            audio_length = get_audio_size(item["audio_file"])
             text_lenght = len(item["text"])
             item["audio_length"] = audio_length
             item["text_length"] = text_lenght

TTS/tts/layers/xtts/trainer/gpt_trainer.py

@@ -321,6 +321,9 @@ class GPTTrainer(BaseTTS):
     def on_train_epoch_start(self, trainer):
         trainer.model.eval()  # the whole model to eval
         # put gpt model in training mode
+        if hasattr(trainer.model, "module") and hasattr(trainer.model.module, "xtts"):
+            trainer.model.module.xtts.gpt.train()
+        else:
             trainer.model.xtts.gpt.train()
 
     def on_init_end(self, trainer):  # pylint: disable=W0613
@@ -387,7 +390,8 @@ class GPTTrainer(BaseTTS):
             else:
                 loader = DataLoader(
                     dataset,
-                    batch_sampler=sampler,
+                    sampler=sampler,
+                    batch_size = config.eval_batch_size if is_eval else config.batch_size,
                     collate_fn=dataset.collate_fn,
                     num_workers=config.num_eval_loader_workers if is_eval else config.num_loader_workers,
                     pin_memory=False,

TTS/tts/layers/xtts/xtts_manager.py

@@ -0,0 +1,34 @@
+import torch
+
+class SpeakerManager():
+    def __init__(self, speaker_file_path=None):
+        self.speakers = torch.load(speaker_file_path)
+
+    @property
+    def name_to_id(self):
+        return self.speakers.keys()
+    
+    @property
+    def num_speakers(self):
+        return len(self.name_to_id)
+    
+    @property
+    def speaker_names(self):
+        return list(self.name_to_id.keys())
+    
+
+class LanguageManager():
+    def __init__(self, config):
+        self.langs = config["languages"]
+
+    @property
+    def name_to_id(self):
+        return self.langs
+    
+    @property
+    def num_languages(self):
+        return len(self.name_to_id)
+    
+    @property
+    def language_names(self):
+        return list(self.name_to_id)

TTS/tts/layers/xtts/zh_num2words.py

@@ -65,7 +65,7 @@ CN_PUNCS_NONSTOP = "＂＃＄％＆＇（）＊＋，－／：；＜＝＞＠［
 CN_PUNCS = CN_PUNCS_STOP + CN_PUNCS_NONSTOP
 
 PUNCS = CN_PUNCS + string.punctuation
-PUNCS_TRANSFORM = str.maketrans(PUNCS, " " * len(PUNCS), "")  # replace puncs with space
+PUNCS_TRANSFORM = str.maketrans(PUNCS, "," * len(PUNCS), "")  # replace puncs with English comma
 
 
 # https://zh.wikipedia.org/wiki/全行和半行

TTS/tts/models/forward_tts.py

@@ -241,7 +241,7 @@ class ForwardTTS(BaseTTS):
         )
 
         self.duration_predictor = DurationPredictor(
-            self.args.hidden_channels + self.embedded_speaker_dim,
+            self.args.hidden_channels,
             self.args.duration_predictor_hidden_channels,
             self.args.duration_predictor_kernel_size,
             self.args.duration_predictor_dropout_p,
@@ -249,7 +249,7 @@ class ForwardTTS(BaseTTS):
 
         if self.args.use_pitch:
             self.pitch_predictor = DurationPredictor(
-                self.args.hidden_channels + self.embedded_speaker_dim,
+                self.args.hidden_channels,
                 self.args.pitch_predictor_hidden_channels,
                 self.args.pitch_predictor_kernel_size,
                 self.args.pitch_predictor_dropout_p,
@@ -263,7 +263,7 @@ class ForwardTTS(BaseTTS):
 
         if self.args.use_energy:
             self.energy_predictor = DurationPredictor(
-                self.args.hidden_channels + self.embedded_speaker_dim,
+                self.args.hidden_channels,
                 self.args.energy_predictor_hidden_channels,
                 self.args.energy_predictor_kernel_size,
                 self.args.energy_predictor_dropout_p,
@@ -299,7 +299,8 @@ class ForwardTTS(BaseTTS):
         if config.use_d_vector_file:
             self.embedded_speaker_dim = config.d_vector_dim
             if self.args.d_vector_dim != self.args.hidden_channels:
-                self.proj_g = nn.Conv1d(self.args.d_vector_dim, self.args.hidden_channels, 1)
+                #self.proj_g = nn.Conv1d(self.args.d_vector_dim, self.args.hidden_channels, 1)
+                self.proj_g = nn.Linear(in_features=self.args.d_vector_dim, out_features=self.args.hidden_channels)
         # init speaker embedding layer
         if config.use_speaker_embedding and not config.use_d_vector_file:
             print(" > Init speaker_embedding layer.")
@@ -403,10 +404,13 @@ class ForwardTTS(BaseTTS):
         # [B, T, C]
         x_emb = self.emb(x)
         # encoder pass
-        o_en = self.encoder(torch.transpose(x_emb, 1, -1), x_mask)
+	#o_en = self.encoder(torch.transpose(x_emb, 1, -1), x_mask)
+        o_en = self.encoder(torch.transpose(x_emb, 1, -1), x_mask, g)
         # speaker conditioning
         # TODO: try different ways of conditioning
         if g is not None: 
+            if hasattr(self, "proj_g"):
+                g = self.proj_g(g.view(g.shape[0], -1)).unsqueeze(-1)            
             o_en = o_en + g
         return o_en, x_mask, g, x_emb
 

TTS/tts/models/xtts.py

@@ -11,6 +11,7 @@ from TTS.tts.layers.xtts.gpt import GPT
 from TTS.tts.layers.xtts.hifigan_decoder import HifiDecoder
 from TTS.tts.layers.xtts.stream_generator import init_stream_support
 from TTS.tts.layers.xtts.tokenizer import VoiceBpeTokenizer, split_sentence
+from TTS.tts.layers.xtts.xtts_manager import SpeakerManager, LanguageManager
 from TTS.tts.models.base_tts import BaseTTS
 from TTS.utils.io import load_fsspec
 
@@ -378,7 +379,7 @@ class Xtts(BaseTTS):
 
         return gpt_cond_latents, speaker_embedding
 
-    def synthesize(self, text, config, speaker_wav, language, **kwargs):
+    def synthesize(self, text, config, speaker_wav, language, speaker_id=None, **kwargs):
         """Synthesize speech with the given input text.
 
         Args:
@@ -393,12 +394,6 @@ class Xtts(BaseTTS):
             `text_input` as text token IDs after tokenizer, `voice_samples` as samples used for cloning, `conditioning_latents`
             as latents used at inference.
 
-        """
-        return self.inference_with_config(text, config, ref_audio_path=speaker_wav, language=language, **kwargs)
-
-    def inference_with_config(self, text, config, ref_audio_path, language, **kwargs):
-        """
-        inference with config
         """
         assert (
             "zh-cn" if language == "zh" else language in self.config.languages
@@ -410,13 +405,18 @@ class Xtts(BaseTTS):
             "repetition_penalty": config.repetition_penalty,
             "top_k": config.top_k,
             "top_p": config.top_p,
+        }
+        settings.update(kwargs)  # allow overriding of preset settings with kwargs
+        if speaker_id is not None:
+            gpt_cond_latent, speaker_embedding = self.speaker_manager.speakers[speaker_id].values()
+            return self.inference(text, language, gpt_cond_latent, speaker_embedding, **settings)
+        settings.update({
             "gpt_cond_len": config.gpt_cond_len,
             "gpt_cond_chunk_len": config.gpt_cond_chunk_len,
             "max_ref_len": config.max_ref_len,
             "sound_norm_refs": config.sound_norm_refs,
-        }
-        settings.update(kwargs)  # allow overriding of preset settings with kwargs
-        return self.full_inference(text, ref_audio_path, language, **settings)
+        })
+        return self.full_inference(text, speaker_wav, language, **settings)
 
     @torch.inference_mode()
     def full_inference(
@@ -520,6 +520,8 @@ class Xtts(BaseTTS):
     ):
         language = language.split("-")[0]  # remove the country code
         length_scale = 1.0 / max(speed, 0.05)
+        gpt_cond_latent = gpt_cond_latent.to(self.device)
+        speaker_embedding = speaker_embedding.to(self.device)
         if enable_text_splitting:
             text = split_sentence(text, language, self.tokenizer.char_limits[language])
         else:
@@ -628,6 +630,8 @@ class Xtts(BaseTTS):
     ):
         language = language.split("-")[0]  # remove the country code
         length_scale = 1.0 / max(speed, 0.05)
+        gpt_cond_latent = gpt_cond_latent.to(self.device)
+        speaker_embedding = speaker_embedding.to(self.device)
         if enable_text_splitting:
             text = split_sentence(text, language, self.tokenizer.char_limits[language])
         else:
@@ -733,6 +737,7 @@ class Xtts(BaseTTS):
         eval=True,
         strict=True,
         use_deepspeed=False,
+        speaker_file_path=None,
     ):
         """
         Loads a checkpoint from disk and initializes the model's state and tokenizer.
@@ -752,6 +757,14 @@ class Xtts(BaseTTS):
         model_path = checkpoint_path or os.path.join(checkpoint_dir, "model.pth")
         vocab_path = vocab_path or os.path.join(checkpoint_dir, "vocab.json")
 
+        if speaker_file_path is None and checkpoint_dir is not None:
+            speaker_file_path = os.path.join(checkpoint_dir, "speakers_xtts.pth")
+
+        self.language_manager = LanguageManager(config)
+        self.speaker_manager = None
+        if speaker_file_path is not None and os.path.exists(speaker_file_path):
+            self.speaker_manager = SpeakerManager(speaker_file_path)
+
         if os.path.exists(vocab_path):
             self.tokenizer = VoiceBpeTokenizer(vocab_file=vocab_path)
 

TTS/utils/generic_utils.py

@@ -36,9 +36,7 @@ def get_git_branch():
         current.replace("* ", "")
     except subprocess.CalledProcessError:
         current = "inside_docker"
-    except FileNotFoundError:
-        current = "unknown"
-    except StopIteration:
+    except (FileNotFoundError, StopIteration) as e:
         current = "unknown"
     return current
 

TTS/utils/manage.py

@@ -11,7 +11,7 @@ import fsspec
 import requests
 from tqdm import tqdm
 
-from TTS.config import load_config
+from TTS.config import load_config, read_json_with_comments
 from TTS.utils.generic_utils import get_user_data_dir
 
 LICENSE_URLS = {
@@ -65,30 +65,7 @@ class ModelManager(object):
         Args:
             file_path (str): path to .models.json.
         """
-        with open(file_path, "r", encoding="utf-8") as json_file:
-            self.models_dict = json.load(json_file)
-
-    def add_cs_api_models(self, model_list: List[str]):
-        """Add list of Coqui Studio model names that are returned from the api
-
-        Each has the following format `<coqui_studio_model>/en/<speaker_name>/<coqui_studio_model>`
-        """
-
-        def _add_model(model_name: str):
-            if not "coqui_studio" in model_name:
-                return
-            model_type, lang, dataset, model = model_name.split("/")
-            if model_type not in self.models_dict:
-                self.models_dict[model_type] = {}
-            if lang not in self.models_dict[model_type]:
-                self.models_dict[model_type][lang] = {}
-            if dataset not in self.models_dict[model_type][lang]:
-                self.models_dict[model_type][lang][dataset] = {}
-            if model not in self.models_dict[model_type][lang][dataset]:
-                self.models_dict[model_type][lang][dataset][model] = {}
-
-        for model_name in model_list:
-            _add_model(model_name)
+        self.models_dict = read_json_with_comments(file_path)
 
     def _list_models(self, model_type, model_count=0):
         if self.verbose:
@@ -315,6 +292,7 @@ class ModelManager(object):
                     f"https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/{model_version}/config.json",
                     f"https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/{model_version}/vocab.json",
                     f"https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/{model_version}/hash.md5",
+                    f"https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/{model_version}/speakers_xtts.pth",
                 ],
             }
         else:
@@ -332,9 +310,9 @@ class ModelManager(object):
     def ask_tos(model_full_path):
         """Ask the user to agree to the terms of service"""
         tos_path = os.path.join(model_full_path, "tos_agreed.txt")
-        print(" > You must agree to the terms of service to use this model.")
-        print(" | > Please see the terms of service at https://coqui.ai/cpml.txt")
-        print(' | > "I have read, understood and agreed to the Terms and Conditions." - [y/n]')
+        print(" > You must confirm the following:")
+        print(' | > "I have purchased a commercial license from Coqui: licensing@coqui.ai"')
+        print(' | > "Otherwise, I agree to the terms of the non-commercial CPML: https://coqui.ai/cpml" - [y/n]')
         answer = input(" | | > ")
         if answer.lower() == "y":
             with open(tos_path, "w", encoding="utf-8") as f:

TTS/utils/synthesizer.py

@@ -305,7 +305,7 @@ class Synthesizer(nn.Module):
         speaker_embedding = None
         speaker_id = None
         if self.tts_speakers_file or hasattr(self.tts_model.speaker_manager, "name_to_id"):
-            if speaker_name and isinstance(speaker_name, str):
+            if speaker_name and isinstance(speaker_name, str) and not self.tts_config.model == "xtts":
                 if self.tts_config.use_d_vector_file:
                     # get the average speaker embedding from the saved d_vectors.
                     speaker_embedding = self.tts_model.speaker_manager.get_mean_embedding(
@@ -335,7 +335,9 @@ class Synthesizer(nn.Module):
         # handle multi-lingual
         language_id = None
         if self.tts_languages_file or (
-            hasattr(self.tts_model, "language_manager") and self.tts_model.language_manager is not None
+            hasattr(self.tts_model, "language_manager") 
+            and self.tts_model.language_manager is not None
+            and not self.tts_config.model == "xtts"
         ):
             if len(self.tts_model.language_manager.name_to_id) == 1:
                 language_id = list(self.tts_model.language_manager.name_to_id.values())[0]
@@ -366,6 +368,7 @@ class Synthesizer(nn.Module):
         if (
             speaker_wav is not None
             and self.tts_model.speaker_manager is not None
+            and hasattr(self.tts_model.speaker_manager, "encoder_ap")
             and self.tts_model.speaker_manager.encoder_ap is not None
         ):
             speaker_embedding = self.tts_model.speaker_manager.compute_embedding_from_clip(speaker_wav)

docs/source/configuration.md

@@ -56,4 +56,4 @@ ModelConfig()
 
 In the example above, ```ModelConfig()``` is the final configuration that the model receives and it has all the fields necessary for the model.
 
-We host pre-defined model configurations under ```TTS/<model_class>/configs/```.Although we recommend a unified config class, you can decompose it as you like as for your custom models as long as all the fields for the trainer, model, and inference APIs are provided.
+We host pre-defined model configurations under ```TTS/<model_class>/configs/```. Although we recommend a unified config class, you can decompose it as you like as for your custom models as long as all the fields for the trainer, model, and inference APIs are provided.

docs/source/finetuning.md

@@ -21,7 +21,7 @@ them and fine-tune it for your own dataset. This will help you in two main ways:
     Fine-tuning comes to the rescue in this case. You can take one of our pre-trained models and fine-tune it on your own
     speech dataset and achieve reasonable results with only a couple of hours of data.
 
-    However, note that, fine-tuning does not ensure great results. The model performance is still depends on the
+    However, note that, fine-tuning does not ensure great results. The model performance still depends on the
     {ref}`dataset quality <what_makes_a_good_dataset>` and the hyper-parameters you choose for fine-tuning. Therefore,
     it still takes a bit of tinkering.
 
@@ -41,7 +41,7 @@ them and fine-tune it for your own dataset. This will help you in two main ways:
     tts --list_models
     ```
 
-    The command above lists the the models in a naming format as ```<model_type>/<language>/<dataset>/<model_name>```.
+    The command above lists the models in a naming format as ```<model_type>/<language>/<dataset>/<model_name>```.
 
     Or you can manually check the `.model.json` file in the project directory.
 

docs/source/formatting_your_dataset.md

@@ -7,7 +7,7 @@ If you have a single audio file and you need to split it into clips, there are d
 
 It is also important to use a lossless audio file format to prevent compression artifacts. We recommend using `wav` file format.
 
-Let's assume you created the audio clips and their transcription. You can collect all your clips under a folder. Let's call this folder `wavs`.
+Let's assume you created the audio clips and their transcription. You can collect all your clips in a folder. Let's call this folder `wavs`.
 
 ```
 /wavs
@@ -17,7 +17,7 @@ Let's assume you created the audio clips and their transcription. You can collec
   ...
 ```
 
-You can either create separate transcription files for each clip or create a text file that maps each audio clip to its transcription. In this file, each column must be delimitered by a special character separating the audio file name, the transcription and the normalized transcription. And make sure that the delimiter is not used in the transcription text.
+You can either create separate transcription files for each clip or create a text file that maps each audio clip to its transcription. In this file, each column must be delimited by a special character separating the audio file name, the transcription and the normalized transcription. And make sure that the delimiter is not used in the transcription text.
 
 We recommend the following format delimited by `|`. In the following example, `audio1`, `audio2` refer to files `audio1.wav`, `audio2.wav` etc.
 
@@ -55,7 +55,7 @@ For more info about dataset qualities and properties check our [post](https://gi
 
 After you collect and format your dataset, you need to check two things. Whether you need a `formatter` and a `text_cleaner`. The `formatter` loads the text file (created above) as a list and the `text_cleaner` performs a sequence of text normalization operations that converts the raw text into the spoken representation (e.g. converting numbers to text, acronyms, and symbols to the spoken format).
 
-If you use a different dataset format then the LJSpeech or the other public datasets that 🐸TTS supports, then you need to write your own `formatter`.
+If you use a different dataset format than the LJSpeech or the other public datasets that 🐸TTS supports, then you need to write your own `formatter`.
 
 If your dataset is in a new language or it needs special normalization steps, then you need a new `text_cleaner`.
 

docs/source/implementing_a_new_language_frontend.md

@@ -2,7 +2,7 @@
 
 - Language frontends are located under `TTS.tts.utils.text`
 - Each special language has a separate folder.
-- Each folder containst all the utilities for processing the text input.
+- Each folder contains all the utilities for processing the text input.
 - `TTS.tts.utils.text.phonemizers` contains the main phonemizer for a language. This is the class that uses the utilities
 from the previous step and used to convert the text to phonemes or graphemes for the model.
 - After you implement your phonemizer, you need to add it to the `TTS/tts/utils/text/phonemizers/__init__.py` to be able to

docs/source/implementing_a_new_model.md

@@ -145,7 +145,7 @@ class MyModel(BaseTTS):
         Args:
             ap (AudioProcessor): audio processor used at training.
             batch (Dict): Model inputs used at the previous training step.
-            outputs (Dict): Model outputs generated at the previoud training step.
+            outputs (Dict): Model outputs generated at the previous training step.
 
         Returns:
             Tuple[Dict, np.ndarray]: training plots and output waveform.
@@ -183,7 +183,7 @@ class MyModel(BaseTTS):
         ...
 
     def get_optimizer(self) -> Union["Optimizer", List["Optimizer"]]:
-        """Setup an return optimizer or optimizers."""
+        """Setup a return optimizer or optimizers."""
         pass
 
     def get_lr(self) -> Union[float, List[float]]:

docs/source/inference.md

@@ -172,48 +172,6 @@ tts.tts_with_vc_to_file(
 )
 ```
 
-#### Example text to speech using [🐸Coqui Studio](https://coqui.ai) models.
-
-You can use all of your available speakers in the studio.
-[🐸Coqui Studio](https://coqui.ai) API token is required. You can get it from the [account page](https://coqui.ai/account).
-You should set the `COQUI_STUDIO_TOKEN` environment variable to use the API token.
-
-```python
-# If you have a valid API token set you will see the studio speakers as separate models in the list.
-# The name format is coqui_studio/en/<studio_speaker_name>/coqui_studio
-models = TTS().list_models()
-# Init TTS with the target studio speaker
-tts = TTS(model_name="coqui_studio/en/Torcull Diarmuid/coqui_studio", progress_bar=False)
-# Run TTS
-tts.tts_to_file(text="This is a test.", file_path=OUTPUT_PATH)
-# Run TTS with emotion and speed control
-tts.tts_to_file(text="This is a test.", file_path=OUTPUT_PATH, emotion="Happy", speed=1.5)
-```
-
-If you just need 🐸 Coqui Studio speakers, you can use `CS_API`. It is a wrapper around the 🐸 Coqui Studio API.
-
-```python
-from TTS.api import CS_API
-
-# Init 🐸 Coqui Studio API
-# you can either set the API token as an environment variable `COQUI_STUDIO_TOKEN` or pass it as an argument.
-
-# XTTS - Best quality and life-like speech in multiple languages. See https://docs.coqui.ai/reference/samples_xtts_create for supported languages.
-api = CS_API(api_token=<token>, model="XTTS")
-api.speakers  # all the speakers are available with all the models.
-api.list_speakers()
-api.list_voices()
-wav, sample_rate = api.tts(text="This is a test.", speaker=api.speakers[0].name, emotion="Happy", language="en", speed=1.5)
-
-# V1 - Fast and lightweight TTS in EN with emotion control.
-api = CS_API(api_token=<token>, model="V1")
-api.speakers
-api.emotions  # emotions are only for the V1 model.
-api.list_speakers()
-api.list_voices()
-wav, sample_rate = api.tts(text="This is a test.", speaker=api.speakers[0].name, emotion="Happy", speed=1.5)
-```
-
 #### Example text to speech using **Fairseq models in ~1100 languages** 🤯.
 For these models use the following name format: `tts_models/<lang-iso_code>/fairseq/vits`.
 

docs/source/marytts.md

@@ -2,13 +2,13 @@
 
 ## What is Mary-TTS?
 
-[Mary (Modular Architecture for Research in sYynthesis) Text-to-Speech](http://mary.dfki.de/) is an open-source (GNU LGPL license), multilingual Text-to-Speech Synthesis platform written in Java. It was originally developed as a collaborative project of [DFKI’s](http://www.dfki.de/web) Language Technology Lab and the [Institute of Phonetics](http://www.coli.uni-saarland.de/groups/WB/Phonetics/) at Saarland University, Germany. It is now maintained by the Multimodal Speech Processing Group in the [Cluster of Excellence MMCI](https://www.mmci.uni-saarland.de/) and DFKI.
+[Mary (Modular Architecture for Research in sYnthesis) Text-to-Speech](http://mary.dfki.de/) is an open-source (GNU LGPL license), multilingual Text-to-Speech Synthesis platform written in Java. It was originally developed as a collaborative project of [DFKI’s](http://www.dfki.de/web) Language Technology Lab and the [Institute of Phonetics](http://www.coli.uni-saarland.de/groups/WB/Phonetics/) at Saarland University, Germany. It is now maintained by the Multimodal Speech Processing Group in the [Cluster of Excellence MMCI](https://www.mmci.uni-saarland.de/) and DFKI.
 MaryTTS has been around for a very! long time. Version 3.0 even dates back to 2006, long before Deep Learning was a broadly known term and the last official release was version 5.2 in 2016.
 You can check out this OpenVoice-Tech page to learn more: https://openvoice-tech.net/index.php/MaryTTS
 
 ## Why Mary-TTS compatibility is relevant
 
-Due to it's open-source nature, relatively high quality voices and fast synthetization speed Mary-TTS was a popular choice in the past and many tools implemented API support over the years like screen-readers (NVDA + SpeechHub), smart-home HUBs (openHAB, Home Assistant) or voice assistants (Rhasspy, Mycroft, SEPIA). A compatibility layer for Coqui-TTS will ensure that these tools can use Coqui as a drop-in replacement and get even better voices right away.
+Due to its open-source nature, relatively high quality voices and fast synthetization speed Mary-TTS was a popular choice in the past and many tools implemented API support over the years like screen-readers (NVDA + SpeechHub), smart-home HUBs (openHAB, Home Assistant) or voice assistants (Rhasspy, Mycroft, SEPIA). A compatibility layer for Coqui-TTS will ensure that these tools can use Coqui as a drop-in replacement and get even better voices right away.
 
 ## API and code examples
 

docs/source/models/tortoise.md

@@ -1,6 +1,6 @@
 # 🐢 Tortoise
 Tortoise is a very expressive TTS system with impressive voice cloning capabilities. It is based on an GPT like autogressive acoustic model that converts input
-text to discritized acouistic tokens, a diffusion model that converts these tokens to melspeectrogram frames and a Univnet vocoder to convert the spectrograms to
+text to discritized acoustic tokens, a diffusion model that converts these tokens to melspectrogram frames and a Univnet vocoder to convert the spectrograms to
 the final audio signal. The important downside is that Tortoise is very slow compared to the parallel TTS models like VITS.
 
 Big thanks to 👑[@manmay-nakhashi](https://github.com/manmay-nakhashi) who helped us implement Tortoise in 🐸TTS.

docs/source/models/xtts.md

@@ -21,7 +21,7 @@ a few tricks to make it faster and support streaming inference.
 - Across the board quality improvements.
 
 ### Code
-Current implementation only supports inference.
+Current implementation only supports inference and GPT encoder training.
 
 ### Languages
 As of now, XTTS-v2 supports 16 languages: English (en), Spanish (es), French (fr), German (de), Italian (it), Portuguese (pt), Polish (pl), Turkish (tr), Russian (ru), Dutch (nl), Czech (cs), Arabic (ar), Chinese (zh-cn), Japanese (ja), Hungarian (hu) and Korean (ko).
@@ -36,9 +36,71 @@ Come and join in our 🐸Community. We're active on [Discord](https://discord.gg
 You can also mail us at info@coqui.ai.
 
 ### Inference
+
+#### 🐸TTS Command line
+
+You can check all supported languages with the following command: 
+
+```console
+ tts --model_name tts_models/multilingual/multi-dataset/xtts_v2 \
+    --list_language_idx
+```
+
+You can check all Coqui available speakers with the following command: 
+
+```console
+ tts --model_name tts_models/multilingual/multi-dataset/xtts_v2 \
+    --list_speaker_idx
+```
+
+##### Coqui speakers
+You can do inference using one of the available speakers using the following command:
+
+```console
+ tts --model_name tts_models/multilingual/multi-dataset/xtts_v2 \
+     --text "It took me quite a long time to develop a voice, and now that I have it I'm not going to be silent." \
+     --speaker_idx "Ana Florence" \
+     --language_idx en \
+     --use_cuda true
+```
+
+##### Clone a voice
+You can clone a speaker voice using a single or multiple references:
+
+###### Single reference
+
+```console
+ tts --model_name tts_models/multilingual/multi-dataset/xtts_v2 \
+     --text "Bugün okula gitmek istemiyorum." \
+     --speaker_wav /path/to/target/speaker.wav \
+     --language_idx tr \
+     --use_cuda true
+```
+
+###### Multiple references
+```console
+ tts --model_name tts_models/multilingual/multi-dataset/xtts_v2 \
+     --text "Bugün okula gitmek istemiyorum." \
+     --speaker_wav /path/to/target/speaker.wav /path/to/target/speaker_2.wav /path/to/target/speaker_3.wav \
+     --language_idx tr \
+     --use_cuda true
+```
+or for all wav files in a directory you can use:
+
+```console
+ tts --model_name tts_models/multilingual/multi-dataset/xtts_v2 \
+     --text "Bugün okula gitmek istemiyorum." \
+     --speaker_wav /path/to/target/*.wav \
+     --language_idx tr \
+     --use_cuda true
+```
+
 #### 🐸TTS API
 
-##### Single reference
+##### Clone a voice
+You can clone a speaker voice using a single or multiple references:
+
+###### Single reference
 
 Splits the text into sentences and generates audio for each sentence. The audio files are then concatenated to produce the final audio.
 You can optionally disable sentence splitting for better coherence but more VRAM and possibly hitting models context length limit.
@@ -56,7 +118,7 @@ tts.tts_to_file(text="It took me quite a long time to develop a voice, and now t
                 )
 ```
 
-##### Multiple references
+###### Multiple references
 
 You can pass multiple audio files to the `speaker_wav` argument for better voice cloning.
 
@@ -81,83 +143,54 @@ tts.tts_to_file(text="It took me quite a long time to develop a voice, and now t
                 language="en")
 ```
 
-##### Streaming inference
+##### Coqui speakers
 
-XTTS supports streaming inference. This is useful for real-time applications.
+You can do inference using one of the available speakers using the following code:
 
 ```python
-import os
-import time
-import torch
-import torchaudio
-
-print("Loading model...")
+from TTS.api import TTS
 tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2", gpu=True)
-model = tts.synthesizer.tts_model
 
-print("Computing speaker latents...")
-gpt_cond_latent, speaker_embedding = model.get_conditioning_latents(audio_path=["reference.wav"])
-
-print("Inference...")
-t0 = time.time()
-stream_generator = model.inference_stream(
-    "It took me quite a long time to develop a voice and now that I have it I am not going to be silent.",
-    "en",
-    gpt_cond_latent,
-    speaker_embedding
-)
-
-wav_chuncks = []
-for i, chunk in enumerate(stream_generator):
-    if i == 0:
-        print(f"Time to first chunck: {time.time() - t0}")
-    print(f"Received chunk {i} of audio length {chunk.shape[-1]}")
-    wav_chuncks.append(chunk)
-wav = torch.cat(wav_chuncks, dim=0)
-torchaudio.save("xtts_streaming.wav", wav.squeeze().unsqueeze(0).cpu(), 24000)
+# generate speech by cloning a voice using default settings
+tts.tts_to_file(text="It took me quite a long time to develop a voice, and now that I have it I'm not going to be silent.",
+                file_path="output.wav",
+                speaker="Ana Florence",
+                language="en",
+                split_sentences=True
+                )
 ```
 
-#### 🐸TTS Command line
-
-##### Single reference
-```console
- tts --model_name tts_models/multilingual/multi-dataset/xtts_v2 \
-     --text "Bugün okula gitmek istemiyorum." \
-     --speaker_wav /path/to/target/speaker.wav \
-     --language_idx tr \
-     --use_cuda true
-```
-
-##### Multiple references
-```console
- tts --model_name tts_models/multilingual/multi-dataset/xtts_v2 \
-     --text "Bugün okula gitmek istemiyorum." \
-     --speaker_wav /path/to/target/speaker.wav /path/to/target/speaker_2.wav /path/to/target/speaker_3.wav \
-     --language_idx tr \
-     --use_cuda true
-```
-or for all wav files in a directory you can use:
-
-```console
- tts --model_name tts_models/multilingual/multi-dataset/xtts_v2 \
-     --text "Bugün okula gitmek istemiyorum." \
-     --speaker_wav /path/to/target/*.wav \
-     --language_idx tr \
-     --use_cuda true
-```
 
 #### 🐸TTS Model API
 
 To use the model API, you need to download the model files and pass config and model file paths manually.
 
-##### Calling manually
+#### Manual Inference
 
-If you want to be able to run with `use_deepspeed=True` and **enjoy the speedup**, you need to install deepspeed first.
+If you want to be able to `load_checkpoint` with `use_deepspeed=True` and **enjoy the speedup**, you need to install deepspeed first.
 
 ```console
 pip install deepspeed==0.10.3
 ```
 
+##### inference parameters
+
+- `text`: The text to be synthesized.
+- `language`: The language of the text to be synthesized.
+- `gpt_cond_latent`: The latent vector you get with get_conditioning_latents. (You can cache for faster inference with same speaker)
+- `speaker_embedding`: The speaker embedding you get with get_conditioning_latents. (You can cache for faster inference with same speaker)
+- `temperature`: The softmax temperature of the autoregressive model. Defaults to 0.65.
+- `length_penalty`: A length penalty applied to the autoregressive decoder. Higher settings causes the model to produce more terse outputs. Defaults to 1.0.
+- `repetition_penalty`: A penalty that prevents the autoregressive decoder from repeating itself during decoding. Can be used to reduce the incidence of long silences or "uhhhhhhs", etc. Defaults to 2.0.
+- `top_k`: Lower values mean the decoder produces more "likely" (aka boring) outputs. Defaults to 50.
+- `top_p`: Lower values mean the decoder produces more "likely" (aka boring) outputs. Defaults to 0.8.
+- `speed`: The speed rate of the generated audio. Defaults to 1.0. (can produce artifacts if far from 1.0)
+- `enable_text_splitting`: Whether to split the text into sentences and generate audio for each sentence. It allows you to have infinite input length but might loose important context between sentences. Defaults to True.
+
+
+##### Inference
+
+
 ```python
 import os
 import torch

requirements.txt

@@ -17,6 +17,7 @@ pyyaml>=6.0
 fsspec>=2023.6.0 # <= 2023.9.1 makes aux tests fail
 aiohttp>=3.8.1
 packaging>=23.1
+mutagen==1.47.0
 # deps for examples
 flask>=2.0.1
 # deps for inference
@@ -27,7 +28,7 @@ pandas>=1.4,<2.0
 # deps for training
 matplotlib>=3.7.0
 # coqui stack
-trainer>=0.0.32
+trainer>=0.0.36
 # config management
 coqpit>=0.0.16
 # chinese g2p deps

tests/api_tests/test_python_api.py

@@ -1,113 +0,0 @@
-import os
-import unittest
-
-from tests import get_tests_data_path, get_tests_output_path
-from TTS.api import CS_API, TTS
-
-OUTPUT_PATH = os.path.join(get_tests_output_path(), "test_python_api.wav")
-cloning_test_wav_path = os.path.join(get_tests_data_path(), "ljspeech/wavs/LJ001-0028.wav")
-
-
-is_coqui_available = os.environ.get("COQUI_STUDIO_TOKEN")
-
-
-if is_coqui_available:
-
-    class CS_APITest(unittest.TestCase):
-        def test_speakers(self):
-            tts = CS_API()
-            self.assertGreater(len(tts.speakers), 1)
-
-        def test_emotions(self):
-            tts = CS_API()
-            self.assertGreater(len(tts.emotions), 1)
-
-        def test_list_calls(self):
-            tts = CS_API()
-            self.assertGreater(len(tts.list_voices()), 1)
-            self.assertGreater(len(tts.list_speakers()), 1)
-            self.assertGreater(len(tts.list_all_speakers()), 1)
-            self.assertGreater(len(tts.list_speakers_as_tts_models()), 1)
-
-        def test_name_to_speaker(self):
-            tts = CS_API()
-            speaker_name = tts.list_speakers_as_tts_models()[0].split("/")[2]
-            speaker = tts.name_to_speaker(speaker_name)
-            self.assertEqual(speaker.name, speaker_name)
-
-        def test_tts(self):
-            tts = CS_API()
-            wav, sr = tts.tts(text="This is a test.", speaker_name=tts.list_speakers()[0].name)
-            self.assertEqual(sr, 44100)
-            self.assertGreater(len(wav), 1)
-
-    class TTSTest(unittest.TestCase):
-        def test_single_speaker_model(self):
-            tts = TTS(model_name="tts_models/de/thorsten/tacotron2-DDC", progress_bar=False, gpu=False)
-
-            error_raised = False
-            try:
-                tts.tts_to_file(text="Ich bin eine Testnachricht.", speaker="Thorsten", language="de")
-            except ValueError:
-                error_raised = True
-
-            tts.tts_to_file(text="Ich bin eine Testnachricht.", file_path=OUTPUT_PATH)
-
-            self.assertTrue(error_raised)
-            self.assertFalse(tts.is_multi_speaker)
-            self.assertFalse(tts.is_multi_lingual)
-            self.assertIsNone(tts.speakers)
-            self.assertIsNone(tts.languages)
-
-        def test_studio_model(self):
-            tts = TTS(model_name="coqui_studio/en/Zacharie Aimilios/coqui_studio")
-            tts.tts_to_file(text="This is a test.")
-
-            # check speed > 2.0 raises error
-            raised_error = False
-            try:
-                _ = tts.tts(text="This is a test.", speed=4.0, emotion="Sad")  # should raise error with speed > 2.0
-            except ValueError:
-                raised_error = True
-            self.assertTrue(raised_error)
-
-            # check emotion is invalid
-            raised_error = False
-            try:
-                _ = tts.tts(text="This is a test.", speed=2.0, emotion="No Emo")  # should raise error with speed > 2.0
-            except ValueError:
-                raised_error = True
-            self.assertTrue(raised_error)
-
-            # check valid call
-            wav = tts.tts(text="This is a test.", speed=2.0, emotion="Sad")
-            self.assertGreater(len(wav), 0)
-
-        def test_fairseq_model(self):  # pylint: disable=no-self-use
-            tts = TTS(model_name="tts_models/eng/fairseq/vits")
-            tts.tts_to_file(text="This is a test.")
-
-        def test_multi_speaker_multi_lingual_model(self):
-            tts = TTS()
-            tts.load_tts_model_by_name(tts.models[0])  # YourTTS
-            tts.tts_to_file(
-                text="Hello world!", speaker=tts.speakers[0], language=tts.languages[0], file_path=OUTPUT_PATH
-            )
-
-            self.assertTrue(tts.is_multi_speaker)
-            self.assertTrue(tts.is_multi_lingual)
-            self.assertGreater(len(tts.speakers), 1)
-            self.assertGreater(len(tts.languages), 1)
-
-        def test_voice_cloning(self):  # pylint: disable=no-self-use
-            tts = TTS()
-            tts.load_tts_model_by_name("tts_models/multilingual/multi-dataset/your_tts")
-            tts.tts_to_file("Hello world!", speaker_wav=cloning_test_wav_path, language="en", file_path=OUTPUT_PATH)
-
-        def test_voice_conversion(self):  # pylint: disable=no-self-use
-            tts = TTS(model_name="voice_conversion_models/multilingual/vctk/freevc24", progress_bar=False, gpu=False)
-            tts.voice_conversion_to_file(
-                source_wav=cloning_test_wav_path,
-                target_wav=cloning_test_wav_path,
-                file_path=OUTPUT_PATH,
-            )

tests/api_tests/test_synthesize_api.py

@@ -1,25 +0,0 @@
-import os
-
-from tests import get_tests_output_path, run_cli
-
-
-def test_synthesize():
-    """Test synthesize.py with diffent arguments."""
-    output_path = os.path.join(get_tests_output_path(), "output.wav")
-
-    # 🐸 Coqui studio model
-    run_cli(
-        'tts --model_name "coqui_studio/en/Torcull Diarmuid/coqui_studio" '
-        '--text "This is it" '
-        f'--out_path "{output_path}"'
-    )
-
-    # 🐸 Coqui studio model with speed arg.
-    run_cli(
-        'tts --model_name "coqui_studio/en/Torcull Diarmuid/coqui_studio" '
-        '--text "This is it but slow" --speed 0.1'
-        f'--out_path "{output_path}"'
-    )
-
-    # test pipe_out command
-    run_cli(f'tts --text "test." --pipe_out --out_path "{output_path}" | aplay')

tests/data/ljspeech/metadata_flac.csv

@@ -0,0 +1,9 @@
+audio_file|text|transcription|speaker_name
+wavs/LJ001-0001.flac|Printing, in the only sense with which we are at present concerned, differs from most if not from all the arts and crafts represented in the Exhibition|Printing, in the only sense with which we are at present concerned, differs from most if not from all the arts and crafts represented in the Exhibition|ljspeech-0
+wavs/LJ001-0002.flac|in being comparatively modern.|in being comparatively modern.|ljspeech-0
+wavs/LJ001-0003.flac|For although the Chinese took impressions from wood blocks engraved in relief for centuries before the woodcutters of the Netherlands, by a similar process|For although the Chinese took impressions from wood blocks engraved in relief for centuries before the woodcutters of the Netherlands, by a similar process|ljspeech-1
+wavs/LJ001-0004.flac|produced the block books, which were the immediate predecessors of the true printed book,|produced the block books, which were the immediate predecessors of the true printed book,|ljspeech-1
+wavs/LJ001-0005.flac|the invention of movable metal letters in the middle of the fifteenth century may justly be considered as the invention of the art of printing.|the invention of movable metal letters in the middle of the fifteenth century may justly be considered as the invention of the art of printing.|ljspeech-2
+wavs/LJ001-0006.flac|And it is worth mention in passing that, as an example of fine typography,|And it is worth mention in passing that, as an example of fine typography,|ljspeech-2
+wavs/LJ001-0007.flac|the earliest book printed with movable types, the Gutenberg, or "forty-two line Bible" of about 1455,|the earliest book printed with movable types, the Gutenberg, or "forty-two line Bible" of about fourteen fifty-five,|ljspeech-3
+wavs/LJ001-0008.flac|has never been surpassed.|has never been surpassed.|ljspeech-3

tests/data/ljspeech/metadata_mp3.csv

@@ -0,0 +1,9 @@
+audio_file|text|transcription|speaker_name
+wavs/LJ001-0001.mp3|Printing, in the only sense with which we are at present concerned, differs from most if not from all the arts and crafts represented in the Exhibition|Printing, in the only sense with which we are at present concerned, differs from most if not from all the arts and crafts represented in the Exhibition|ljspeech-0
+wavs/LJ001-0002.mp3|in being comparatively modern.|in being comparatively modern.|ljspeech-0
+wavs/LJ001-0003.mp3|For although the Chinese took impressions from wood blocks engraved in relief for centuries before the woodcutters of the Netherlands, by a similar process|For although the Chinese took impressions from wood blocks engraved in relief for centuries before the woodcutters of the Netherlands, by a similar process|ljspeech-1
+wavs/LJ001-0004.mp3|produced the block books, which were the immediate predecessors of the true printed book,|produced the block books, which were the immediate predecessors of the true printed book,|ljspeech-1
+wavs/LJ001-0005.mp3|the invention of movable metal letters in the middle of the fifteenth century may justly be considered as the invention of the art of printing.|the invention of movable metal letters in the middle of the fifteenth century may justly be considered as the invention of the art of printing.|ljspeech-2
+wavs/LJ001-0006.mp3|And it is worth mention in passing that, as an example of fine typography,|And it is worth mention in passing that, as an example of fine typography,|ljspeech-2
+wavs/LJ001-0007.mp3|the earliest book printed with movable types, the Gutenberg, or "forty-two line Bible" of about 1455,|the earliest book printed with movable types, the Gutenberg, or "forty-two line Bible" of about fourteen fifty-five,|ljspeech-3
+wavs/LJ001-0008.mp3|has never been surpassed.|has never been surpassed.|ljspeech-3

tests/data/ljspeech/metadata_wav.csv

@@ -0,0 +1,9 @@
+audio_file|text|transcription|speaker_name
+wavs/LJ001-0001.wav|Printing, in the only sense with which we are at present concerned, differs from most if not from all the arts and crafts represented in the Exhibition|Printing, in the only sense with which we are at present concerned, differs from most if not from all the arts and crafts represented in the Exhibition|ljspeech-0
+wavs/LJ001-0002.wav|in being comparatively modern.|in being comparatively modern.|ljspeech-0
+wavs/LJ001-0003.wav|For although the Chinese took impressions from wood blocks engraved in relief for centuries before the woodcutters of the Netherlands, by a similar process|For although the Chinese took impressions from wood blocks engraved in relief for centuries before the woodcutters of the Netherlands, by a similar process|ljspeech-1
+wavs/LJ001-0004.wav|produced the block books, which were the immediate predecessors of the true printed book,|produced the block books, which were the immediate predecessors of the true printed book,|ljspeech-1
+wavs/LJ001-0005.wav|the invention of movable metal letters in the middle of the fifteenth century may justly be considered as the invention of the art of printing.|the invention of movable metal letters in the middle of the fifteenth century may justly be considered as the invention of the art of printing.|ljspeech-2
+wavs/LJ001-0006.wav|And it is worth mention in passing that, as an example of fine typography,|And it is worth mention in passing that, as an example of fine typography,|ljspeech-2
+wavs/LJ001-0007.wav|the earliest book printed with movable types, the Gutenberg, or "forty-two line Bible" of about 1455,|the earliest book printed with movable types, the Gutenberg, or "forty-two line Bible" of about fourteen fifty-five,|ljspeech-3
+wavs/LJ001-0008.wav|has never been surpassed.|has never been surpassed.|ljspeech-3