Add version 3.0.0

robinhad · Sep 14, 2022 · a163565 · a163565
1 parent 7af6095
commit a163565
Show file tree

Hide file tree

Showing 3 changed files with 47 additions and 20 deletions.
diff --git a/README.md b/README.md
@@ -18,8 +18,10 @@ Link to source code and models -> [https://github.com/robinhad/ukrainian-tts](ht
 
 Code is licensed under `MIT License`, models are under `GNU GPL v3 License`. 
 # Support
-If you like my work, please support -> [https://send.monobank.ua/jar/48iHq4xAXm](https://send.monobank.ua/jar/48iHq4xAXm)
-# Example
+If you like my work, please support -> [https://send.monobank.ua/jar/48iHq4xAXm](https://send.monobank.ua/jar/48iHq4xAXm)  
+For collaboration and question please contact me here: [Telegram https://t.me/robinhad](https://t.me/robinhad) [Twitter https://twitter.com/robinhad](https://twitter.com/robinhad)  
+You're welcome to join UA Speech Recognition and Synthesis community: [Telegram https://t.me/speech_recognition_uk](https://t.me/speech_recognition_uk)
+# Examples
 
 `Mykyta (male)`:
 
@@ -53,7 +55,8 @@ tts-server --model_path path/to/model.pth \
 # Attribution 🤝
 
 - Model training - [Yurii Paniv @robinhad](https://github.com/robinhad)   
-- Mykyta, Olena and Lada dataset - [Yehor Smoliakov @egorsmkv](https://github.com/egorsmkv)   
+- Mykyta, Olena, Lada, Dmytro, Olha dataset - [Yehor Smoliakov @egorsmkv](https://github.com/egorsmkv)   
+- Dmytro voice - [Dmytro Chaplynskyi @dchaplinsky](https://github.com/dchaplinsky)  
 - Silence cutting using [HMM-GMM](https://github.com/proger/uk) - [Volodymyr Kyrylov @proger](https://github.com/proger)  
 - Autostress (with dictionary) using [ukrainian-word-stress](https://github.com/lang-uk/ukrainian-word-stress) - [Oleksiy Syvokon @asivokon](https://github.com/asivokon)    
 - Autostress (with model) using [ukrainian-accentor](https://github.com/egorsmkv/ukrainian-accentor) - [Bohdan Mykhailenko @NeonBohdan](https://github.com/NeonBohdan) + [Yehor Smoliakov @egorsmkv](https://github.com/egorsmkv)    
diff --git a/app.py b/app.py
@@ -20,6 +20,8 @@ class VoiceOption(Enum):
     Olena = "Олена (жіночий) 👩"
     Mykyta = "Микита (чоловічий) 👨"
     Lada = "Лада (жіночий) 👩"
+    Dmytro = "Дмитро (чоловічий) 👩"
+    Olga = "Ольга (жіночий) 👩"
 
 
 def download(url, file_name):
@@ -33,7 +35,7 @@ def download(url, file_name):
 
 
 print("downloading uk/mykyta/vits-tts")
-release_number = "v3.0.0-alpha"
+release_number = "v3.0.0"
 model_link = f"https://github.com/robinhad/ukrainian-tts/releases/download/{release_number}/model-inference.pth"
 config_link = f"https://github.com/robinhad/ukrainian-tts/releases/download/{release_number}/config.json"
 speakers_link = f"https://github.com/robinhad/ukrainian-tts/releases/download/{release_number}/speakers.pth"
@@ -71,11 +73,14 @@ def tts(text: str, voice: str, stress: str):
     autostress_with_model = (
         True if stress == StressOption.AutomaticStressWithModel.value else False
     )
-    speaker_name = "mykyta" 
-    if voice == VoiceOption.Olena.value:
-        speaker_name = "olena"
-    elif voice == VoiceOption.Lada.value:
-        speaker_name = "lada"
+    voice_mapping = {
+        VoiceOption.Olena.value: "olena",
+        VoiceOption.Mykyta.value: "mykyta",
+        VoiceOption.Lada.value: "lada",
+        VoiceOption.Dmytro.value: "dmytro",
+        VoiceOption.Olga.value: "olga",
+    }
+    speaker_name = voice_mapping[voice]
     text = preprocess_text(text, autostress_with_model)
     text_limit = 7200
     text = (
@@ -98,23 +103,24 @@ def tts(text: str, voice: str, stress: str):
 iface = gr.Interface(
     fn=tts,
     inputs=[
-        gr.inputs.Textbox(
+        gr.components.Textbox(
             label="Input",
-            default="Введіть, будь ласка, своє р+ечення.",
+            value="Введіть, будь ласка, своє р+ечення.",
         ),
-        gr.inputs.Radio(
+        gr.components.Radio(
             label="Голос",
             choices=[option.value for option in VoiceOption],
-            default=VoiceOption.Olena.value,
+            value=VoiceOption.Olena.value,
         ),
-        gr.inputs.Radio(
+        gr.components.Radio(
             label="Наголоси",
             choices=[option.value for option in StressOption],
+            value=StressOption.AutomaticStress.value
         ),
     ],
     outputs=[
-        gr.outputs.Audio(label="Output"),
-        gr.outputs.Textbox(label="Наголошений текст"),
+        gr.components.Audio(label="Output"),
+        gr.components.Textbox(label="Наголошений текст"),
     ],
     title="🐸💬🇺🇦 - Coqui TTS",
     description="Україномовний🇺🇦 TTS за допомогою Coqui TTS (щоб вручну поставити наголос, використовуйте + перед голосною)",
@@ -132,12 +138,12 @@ def tts(text: str, voice: str, stress: str):
         ],
         [
             "Вв+едіть, будь ласка, св+оє реч+ення.",
-            VoiceOption.Mykyta.value,
+            VoiceOption.Dmytro.value,
             StressOption.AutomaticStress.value,
         ],
         [
             "Привіт, як тебе звати?",
-            VoiceOption.Olena.value,
+            VoiceOption.Olga.value,
             StressOption.AutomaticStress.value,
         ],
         [

diff --git a/config.json b/config.json
@@ -73,7 +73,7 @@
         "griffin_lim_iters": 60,
         "num_mels": 80,
         "mel_fmin": 0,
-        "mel_fmax": 8000,
+        "mel_fmax": null,
         "spec_gain": 6.0,
         "do_amp_to_db_linear": true,
         "do_amp_to_db_mel": true,
@@ -158,11 +158,29 @@
             null,
             null
         ],
+        [
+            "\u0425\u0442+\u043e \u0442+\u0438 \u0442\u0430\u043a+\u0438\u0439 +\u0456 +\u044f\u043a \u0442\u0435\u0431+\u0435 \u0437\u0432+\u0430\u0442\u0438?",
+            "dmytro",
+            null,
+            null
+        ],
         [
             "\u041a\u0430\u043c'\u044f\u043d+\u0435\u0446\u044c-\u041f\u043e\u0434+\u0456\u043b\u044c\u0441\u044c\u043a\u0438\u0439 - \u043c+\u0456\u0441\u0442\u043e \u0432 \u0425\u043c\u0435\u043b\u044c\u043d+\u0438\u0446\u044c\u043a\u0456\u0439 +\u043e\u0431\u043b\u0430\u0441\u0442\u0456 \u0423\u043a\u0440\u0430+\u0457\u043d\u0438, \u0446+\u0435\u043d\u0442\u0440 \u041a\u0430\u043c'\u044f\u043d+\u0435\u0446\u044c-\u041f\u043e\u0434+\u0456\u043b\u044c\u0441\u044c\u043a\u043e\u0457 \u043c\u0456\u0441\u044c\u043a+\u043e\u0457 \u043e\u0431'+\u0454\u0434\u043d\u0430\u043d\u043e\u0457 \u0442\u0435\u0440\u0438\u0442\u043e\u0440\u0456+\u0430\u043b\u044c\u043d\u043e\u0457 \u0433\u0440\u043e\u043c+\u0430\u0434\u0438 +\u0456 \u041a\u0430\u043c'\u044f\u043d+\u0435\u0446\u044c-\u041f\u043e\u0434+\u0456\u043b\u044c\u0441\u044c\u043a\u043e\u0433\u043e \u0440\u0430\u0439+\u043e\u043d\u0443.",
             "lada",
             null,
             null
+        ],
+        [
+            "\u041a\u0430\u043c'\u044f\u043d+\u0435\u0446\u044c-\u041f\u043e\u0434+\u0456\u043b\u044c\u0441\u044c\u043a\u0438\u0439 - \u043c+\u0456\u0441\u0442\u043e \u0432 \u0425\u043c\u0435\u043b\u044c\u043d+\u0438\u0446\u044c\u043a\u0456\u0439 +\u043e\u0431\u043b\u0430\u0441\u0442\u0456 \u0423\u043a\u0440\u0430+\u0457\u043d\u0438, \u0446+\u0435\u043d\u0442\u0440 \u041a\u0430\u043c'\u044f\u043d+\u0435\u0446\u044c-\u041f\u043e\u0434+\u0456\u043b\u044c\u0441\u044c\u043a\u043e\u0457 \u043c\u0456\u0441\u044c\u043a+\u043e\u0457 \u043e\u0431'+\u0454\u0434\u043d\u0430\u043d\u043e\u0457 \u0442\u0435\u0440\u0438\u0442\u043e\u0440\u0456+\u0430\u043b\u044c\u043d\u043e\u0457 \u0433\u0440\u043e\u043c+\u0430\u0434\u0438 +\u0456 \u041a\u0430\u043c'\u044f\u043d+\u0435\u0446\u044c-\u041f\u043e\u0434+\u0456\u043b\u044c\u0441\u044c\u043a\u043e\u0433\u043e \u0440\u0430\u0439+\u043e\u043d\u0443.",
+            "dmytro",
+            null,
+            null
+        ],
+        [
+            "\u041a\u0430\u043c'\u044f\u043d+\u0435\u0446\u044c-\u041f\u043e\u0434+\u0456\u043b\u044c\u0441\u044c\u043a\u0438\u0439 - \u043c+\u0456\u0441\u0442\u043e \u0432 \u0425\u043c\u0435\u043b\u044c\u043d+\u0438\u0446\u044c\u043a\u0456\u0439 +\u043e\u0431\u043b\u0430\u0441\u0442\u0456 \u0423\u043a\u0440\u0430+\u0457\u043d\u0438, \u0446+\u0435\u043d\u0442\u0440 \u041a\u0430\u043c'\u044f\u043d+\u0435\u0446\u044c-\u041f\u043e\u0434+\u0456\u043b\u044c\u0441\u044c\u043a\u043e\u0457 \u043c\u0456\u0441\u044c\u043a+\u043e\u0457 \u043e\u0431'+\u0454\u0434\u043d\u0430\u043d\u043e\u0457 \u0442\u0435\u0440\u0438\u0442\u043e\u0440\u0456+\u0430\u043b\u044c\u043d\u043e\u0457 \u0433\u0440\u043e\u043c+\u0430\u0434\u0438 +\u0456 \u041a\u0430\u043c'\u044f\u043d+\u0435\u0446\u044c-\u041f\u043e\u0434+\u0456\u043b\u044c\u0441\u044c\u043a\u043e\u0433\u043e \u0440\u0430\u0439+\u043e\u043d\u0443.",
+            "olga",
+            null,
+            null
         ]
     ],
     "eval_split_max_size": null,
@@ -243,7 +261,7 @@
         "init_discriminator": true,
         "use_spectral_norm_disriminator": false,
         "use_speaker_embedding": true,
-        "num_speakers": 3,
+        "num_speakers": 5,
         "speakers_file": "speakers.pth",
         "d_vector_file": null,
         "speaker_embedding_channels": 256,