Import spaCy languages dynamically in XTTS Tokenizer.py

TTS/.models.json

@@ -46,7 +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://coqui.gateway.scarf.sh/hf/text_2.pt",
+                        "https://coqui.gateway.scarf.sh/hf/bark/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"

TTS/server/server.py

@@ -191,10 +191,19 @@ def details():
 @app.route("/api/tts", methods=["GET", "POST"])
 def tts():
     with lock:
-        text = request.headers.get("text") or request.values.get("text", "")
-        speaker_idx = request.headers.get("speaker-id") or request.values.get("speaker_id", "")
-        language_idx = request.headers.get("language-id") or request.values.get("language_id", "")
-        style_wav = request.headers.get("style-wav") or request.values.get("style_wav", "")
+        try:
+            data = request.get_json()
+            text = data.get("text", "")
+            speaker_idx = data.get("speaker-id", "")
+            language_idx = data.get("language-id", "")
+            style_wav = data.get("style-wav", "")
+        except:
+            # Fallback to headers and form data if JSON data is not present
+            text = request.headers.get("text") or request.values.get("text", "")
+            speaker_idx = request.headers.get("speaker-id") or request.values.get("speaker_id", "")
+            language_idx = request.headers.get("language-id") or request.values.get("language_id", "")
+            style_wav = request.headers.get("style-wav") or request.values.get("style_wav", "")
+
         style_wav = style_wav_uri_to_dict(style_wav)
 
         print(f" > Model input: {text}")

TTS/tts/layers/xtts/tokenizer.py

@@ -8,29 +8,37 @@
 from hangul_romanize import Transliter
 from hangul_romanize.rule import academic
 from num2words import num2words
-from spacy.lang.ar import Arabic
 from spacy.lang.en import English
-from spacy.lang.es import Spanish
-from spacy.lang.ja import Japanese
-from spacy.lang.zh import Chinese
+from spacy.util import get_lang_class
+
 from tokenizers import Tokenizer
 
 from TTS.tts.layers.xtts.zh_num2words import TextNorm as zh_num2words
 
+import spacy
+
+# These 2 functions are to verify that any language can be instantiated
+def get_spacy_available_langs():
+    from pathlib import Path
+    spacy_path = Path(spacy.__file__.replace('__init__.py',''))
+    spacy_langs = spacy_path / 'lang'
+    SPACY_LANGS = [str(x).split('/')[-1] for x in spacy_langs.iterdir() if x.is_dir() and str(x).split('/')[-1] != '__pycache__']
+    print("Available languages in Spacy:", SPACY_LANGS)
+    return SPACY_LANGS
+def get_all_spacy_langs():
+    SPACY_LANGS = get_spacy_available_langs()
+    spacy_lang_instances = []
+    for lang in SPACY_LANGS:
+        spacy_lang_instances.append(get_spacy_lang(lang))
 
 def get_spacy_lang(lang):
-    if lang == "zh":
-        return Chinese()
-    elif lang == "ja":
-        return Japanese()
-    elif lang == "ar":
-        return Arabic()
-    elif lang == "es":
-        return Spanish()
-    else:
-        # For most languages, Enlish does the job
-        return English()
-
+    try:
+        lang_model = get_lang_class(lang)()
+    except ImportError as e:
+        print("Error", e)
+        # Fallback to English if the language model is not available
+        lang_model = English()
+    return lang_model
 
 def split_sentence(text, lang, text_split_length=250):
     """Preprocess the input text"""

TTS/utils/manage.py

@@ -260,10 +260,13 @@ def set_model_url(model_item: Dict):
     def _set_model_item(self, model_name):
         # fetch model info from the dict
         if "fairseq" in model_name:
-            model_type = "tts_models"
-            lang = model_name.split("/")[1]
+            split = model_name.split("/")
+            model_type = split[0]
+            lang = split[1]
+            dataset = split[2]
+            model = split[3]
             model_item = {
-                "model_type": "tts_models",
+                "model_type": model_type,
                 "license": "CC BY-NC 4.0",
                 "default_vocoder": None,
                 "author": "fairseq",

TTS/utils/synthesizer.py

@@ -264,6 +264,7 @@ def tts(
         style_text=None,
         reference_wav=None,
         reference_speaker_name=None,
+        verbose: bool = True,
         split_sentences: bool = True,
         **kwargs,
     ) -> List[int]:
@@ -278,6 +279,7 @@ def tts(
             style_text ([type], optional): transcription of style_wav for Capacitron. Defaults to None.
             reference_wav ([type], optional): reference waveform for voice conversion. Defaults to None.
             reference_speaker_name ([type], optional): speaker id of reference waveform. Defaults to None.
+            verbose (bool, optional): print verbose output. Defaults to True.
             split_sentences (bool, optional): split the input text into sentences. Defaults to True.
             **kwargs: additional arguments to pass to the TTS model.
         Returns:
@@ -294,9 +296,11 @@ def tts(
         if text:
             sens = [text]
             if split_sentences:
-                print(" > Text splitted to sentences.")
+                if verbose:
+                    print(" > Text splitted to sentences.")
                 sens = self.split_into_sentences(text)
-            print(sens)
+            if verbose:
+                print(sens)
 
         # handle multi-speaker
         if "voice_dir" in kwargs:
@@ -420,7 +424,8 @@ def tts(
                         self.vocoder_config["audio"]["sample_rate"] / self.tts_model.ap.sample_rate,
                     ]
                     if scale_factor[1] != 1:
-                        print(" > interpolating tts model output.")
+                        if verbose:
+                            print(" > interpolating tts model output.")
                         vocoder_input = interpolate_vocoder_input(scale_factor, vocoder_input)
                     else:
                         vocoder_input = torch.tensor(vocoder_input).unsqueeze(0)  # pylint: disable=not-callable
@@ -484,7 +489,8 @@ def tts(
                     self.vocoder_config["audio"]["sample_rate"] / self.tts_model.ap.sample_rate,
                 ]
                 if scale_factor[1] != 1:
-                    print(" > interpolating tts model output.")
+                    if verbose:
+                        print(" > interpolating tts model output.")
                     vocoder_input = interpolate_vocoder_input(scale_factor, vocoder_input)
                 else:
                     vocoder_input = torch.tensor(vocoder_input).unsqueeze(0)  # pylint: disable=not-callable
@@ -497,9 +503,10 @@ def tts(
                 waveform = waveform.numpy()
             wavs = waveform.squeeze()
 
-        # compute stats
-        process_time = time.time() - start_time
-        audio_time = len(wavs) / self.tts_config.audio["sample_rate"]
-        print(f" > Processing time: {process_time}")
-        print(f" > Real-time factor: {process_time / audio_time}")
+        if verbose:
+            # compute stats
+            process_time = time.time() - start_time
+            audio_time = len(wavs) / self.tts_config.audio["sample_rate"]
+            print(f" > Processing time: {process_time}")
+            print(f" > Real-time factor: {process_time / audio_time}")
         return wavs

requirements.txt

@@ -54,4 +54,10 @@ encodec>=0.1.1
 # deps for XTTS
 unidecode>=1.3.2
 num2words
-spacy[ja]>=3
+spacy[ja]>=3
+
+# spacy thai
+pythainlp>=5.0.1
+
+#spacy vietnamese
+pyvi>=0.1.1