Create app.py

app.py

@@ -0,0 +1,497 @@
+import tempfile
+
+import edge_tts
+import gradio as gr
+from gradio_client import Client
+import pyarabic.araby as araby
+
+language_dict = {
+  "English": {
+    "Jenny": "en-US-JennyNeural",
+    "Guy": "en-US-GuyNeural",
+    "Ana": "en-US-AnaNeural",
+    "Aria": "en-US-AriaNeural",
+    "Christopher": "en-US-ChristopherNeural",
+    "Eric": "en-US-EricNeural",
+    "Michelle": "en-US-MichelleNeural",
+    "Roger": "en-US-RogerNeural",
+    "Natasha": "en-AU-NatashaNeural",
+    "William": "en-AU-WilliamNeural",
+    "Clara": "en-CA-ClaraNeural",
+    "Liam": "en-CA-LiamNeural",
+    "Libby": "en-GB-LibbyNeural",
+    "Maisie": "en-GB-MaisieNeural",
+    "Ryan": "en-GB-RyanNeural",
+    "Sonia": "en-GB-SoniaNeural",
+    "Thomas": "en-GB-ThomasNeural",
+    "Sam": "en-HK-SamNeural",
+    "Yan": "en-HK-YanNeural",
+    "Connor": "en-IE-ConnorNeural",
+    "Emily": "en-IE-EmilyNeural",
+    "Neerja": "en-IN-NeerjaNeural",
+    "Prabhat": "en-IN-PrabhatNeural",
+    "Asilia": "en-KE-AsiliaNeural",
+    "Chilemba": "en-KE-ChilembaNeural",
+    "Abeo": "en-NG-AbeoNeural",
+    "Ezinne": "en-NG-EzinneNeural",
+    "Mitchell": "en-NZ-MitchellNeural",
+    "James": "en-PH-JamesNeural",
+    "Rosa": "en-PH-RosaNeural",
+    "Luna": "en-SG-LunaNeural",
+    "Wayne": "en-SG-WayneNeural",
+    "Elimu": "en-TZ-ElimuNeural",
+    "Imani": "en-TZ-ImaniNeural",
+    "Leah": "en-ZA-LeahNeural",
+    "Luke": "en-ZA-LukeNeural"
+  },
+  "Spanish": {
+    "Elena": "es-AR-ElenaNeural",
+    "Tomas": "es-AR-TomasNeural",
+    "Marcelo": "es-BO-MarceloNeural",
+    "Sofia": "es-BO-SofiaNeural",
+    "Gonzalo": "es-CO-GonzaloNeural",
+    "Salome": "es-CO-SalomeNeural",
+    "Juan": "es-CR-JuanNeural",
+    "Maria": "es-CR-MariaNeural",
+    "Belkys": "es-CU-BelkysNeural",
+    "Emilio": "es-DO-EmilioNeural",
+    "Ramona": "es-DO-RamonaNeural",
+    "Andrea": "es-EC-AndreaNeural",
+    "Luis": "es-EC-LuisNeural",
+    "Alvaro": "es-ES-AlvaroNeural",
+    "Elvira": "es-ES-ElviraNeural",
+    "Teresa": "es-GQ-TeresaNeural",
+    "Andres": "es-GT-AndresNeural",
+    "Marta": "es-GT-MartaNeural",
+    "Carlos": "es-HN-CarlosNeural",
+    "Karla": "es-HN-KarlaNeural",
+    "Federico": "es-NI-FedericoNeural",
+    "Yolanda": "es-NI-YolandaNeural",
+    "Margarita": "es-PA-MargaritaNeural",
+    "Roberto": "es-PA-RobertoNeural",
+    "Alex": "es-PE-AlexNeural",
+    "Camila": "es-PE-CamilaNeural",
+    "Karina": "es-PR-KarinaNeural",
+    "Victor": "es-PR-VictorNeural",
+    "Mario": "es-PY-MarioNeural",
+    "Tania": "es-PY-TaniaNeural",
+    "Lorena": "es-SV-LorenaNeural",
+    "Rodrigo": "es-SV-RodrigoNeural",
+    "Alonso": "es-US-AlonsoNeural",
+    "Paloma": "es-US-PalomaNeural",
+    "Mateo": "es-UY-MateoNeural",
+    "Valentina": "es-UY-ValentinaNeural",
+    "Paola": "es-VE-PaolaNeural",
+    "Sebastian": "es-VE-SebastianNeural"
+  },
+  "Arabic": {
+    "Hamed": "ar-SA-HamedNeural",
+    "Zariyah": "ar-SA-ZariyahNeural",
+    "Fatima": "ar-AE-FatimaNeural",
+    "Hamdan": "ar-AE-HamdanNeural",
+    "Ali": "ar-BH-AliNeural",
+    "Laila": "ar-BH-LailaNeural",
+    "Ismael": "ar-DZ-IsmaelNeural",
+    "Salma": "ar-EG-SalmaNeural",
+    "Shakir": "ar-EG-ShakirNeural",
+    "Bassel": "ar-IQ-BasselNeural",
+    "Rana": "ar-IQ-RanaNeural",
+    "Sana": "ar-JO-SanaNeural",
+    "Taim": "ar-JO-TaimNeural",
+    "Fahed": "ar-KW-FahedNeural",
+    "Noura": "ar-KW-NouraNeural",
+    "Layla": "ar-LB-LaylaNeural",
+    "Rami": "ar-LB-RamiNeural",
+    "Iman": "ar-LY-ImanNeural",
+    "Omar": "ar-LY-OmarNeural",
+    "Jamal": "ar-MA-JamalNeural",
+    "Mouna": "ar-MA-MounaNeural",
+    "Abdullah": "ar-OM-AbdullahNeural",
+    "Aysha": "ar-OM-AyshaNeural",
+    "Amal": "ar-QA-AmalNeural",
+    "Moaz": "ar-QA-MoazNeural",
+    "Amany": "ar-SY-AmanyNeural",
+    "Laith": "ar-SY-LaithNeural",
+    "Hedi": "ar-TN-HediNeural",
+    "Reem": "ar-TN-ReemNeural",
+    "Maryam": "ar-YE-MaryamNeural",
+    "Saleh": "ar-YE-SalehNeural"
+  },
+  "Korean": {
+    "Sun-Hi": "ko-KR-SunHiNeural",
+    "InJoon": "ko-KR-InJoonNeural"
+  },
+  "Thai": {
+    "Premwadee": "th-TH-PremwadeeNeural",
+    "Niwat": "th-TH-NiwatNeural"
+  },
+  "Vietnamese": {
+    "HoaiMy": "vi-VN-HoaiMyNeural",
+    "NamMinh": "vi-VN-NamMinhNeural"
+  },
+  "Japanese": {
+    "Nanami": "ja-JP-NanamiNeural",
+    "Keita": "ja-JP-KeitaNeural"
+  },
+  "French": {
+    "Denise": "fr-FR-DeniseNeural",
+    "Eloise": "fr-FR-EloiseNeural",
+    "Henri": "fr-FR-HenriNeural",
+    "Sylvie": "fr-CA-SylvieNeural",
+    "Antoine": "fr-CA-AntoineNeural",
+    "Jean": "fr-CA-JeanNeural",
+    "Ariane": "fr-CH-ArianeNeural",
+    "Fabrice": "fr-CH-FabriceNeural",
+    "Charline": "fr-BE-CharlineNeural",
+    "Gerard": "fr-BE-GerardNeural"
+  },
+  "Portuguese": {
+    "Francisca": "pt-BR-FranciscaNeural",
+    "Antonio": "pt-BR-AntonioNeural",
+    "Duarte": "pt-PT-DuarteNeural",
+    "Raquel": "pt-PT-RaquelNeural"
+  },
+  "Indonesian": {
+    "Ardi": "id-ID-ArdiNeural",
+    "Gadis": "id-ID-GadisNeural"
+  },
+  "Hebrew": {
+    "Avri": "he-IL-AvriNeural",
+    "Hila": "he-IL-HilaNeural"
+  },
+  "Italian": {
+    "Isabella": "it-IT-IsabellaNeural",
+    "Diego": "it-IT-DiegoNeural",
+    "Elsa": "it-IT-ElsaNeural"
+  },
+  "Dutch": {
+    "Colette": "nl-NL-ColetteNeural",
+    "Fenna": "nl-NL-FennaNeural",
+    "Maarten": "nl-NL-MaartenNeural",
+    "Arnaud": "nl-BE-ArnaudNeural",
+    "Dena": "nl-BE-DenaNeural"
+  },
+  "Malay": {
+    "Osman": "ms-MY-OsmanNeural",
+    "Yasmin": "ms-MY-YasminNeural"
+  },
+  "Norwegian": {
+    "Pernille": "nb-NO-PernilleNeural",
+    "Finn": "nb-NO-FinnNeural"
+  },
+  "Swedish": {
+    "Sofie": "sv-SE-SofieNeural",
+    "Mattias": "sv-SE-MattiasNeural"
+  },
+  "Greek": {
+    "Athina": "el-GR-AthinaNeural",
+    "Nestoras": "el-GR-NestorasNeural"
+  },
+  "German": {
+    "Katja": "de-DE-KatjaNeural",
+    "Amala": "de-DE-AmalaNeural",
+    "Conrad": "de-DE-ConradNeural",
+    "Killian": "de-DE-KillianNeural",
+    "Ingrid": "de-AT-IngridNeural",
+    "Jonas": "de-AT-JonasNeural",
+    "Jan": "de-CH-JanNeural",
+    "Leni": "de-CH-LeniNeural"
+  },
+  "Afrikaans": {
+    "Adri": "af-ZA-AdriNeural",
+    "Willem": "af-ZA-WillemNeural"
+  },
+  "Amharic": {
+    "Ameha": "am-ET-AmehaNeural",
+    "Mekdes": "am-ET-MekdesNeural"
+  },
+  "Azerbaijani": {
+    "Babek": "az-AZ-BabekNeural",
+    "Banu": "az-AZ-BanuNeural"
+  },
+  "Bulgarian": {
+    "Borislav": "bg-BG-BorislavNeural",
+    "Kalina": "bg-BG-KalinaNeural"
+  },
+  "Bengali": {
+    "Nabanita": "bn-BD-NabanitaNeural",
+    "Pradeep": "bn-BD-PradeepNeural",
+    "Bashkar": "bn-IN-BashkarNeural",
+    "Tanishaa": "bn-IN-TanishaaNeural"
+  },
+  "Bosnian": {
+    "Goran": "bs-BA-GoranNeural",
+    "Vesna": "bs-BA-VesnaNeural"
+  },
+  "Catalan": {
+    "Joana": "ca-ES-JoanaNeural",
+    "Enric": "ca-ES-EnricNeural"
+  },
+  "Czech": {
+    "Antonin": "cs-CZ-AntoninNeural",
+    "Vlasta": "cs-CZ-VlastaNeural"
+  },
+  "Welsh": {
+    "Aled": "cy-GB-AledNeural",
+    "Nia": "cy-GB-NiaNeural"
+  },
+  "Danish": {
+    "Christel": "da-DK-ChristelNeural",
+    "Jeppe": "da-DK-JeppeNeural"
+  },
+  "Estonian": {
+    "Anu": "et-EE-AnuNeural",
+    "Kert": "et-EE-KertNeural"
+  },
+  "Persian": {
+    "Dilara": "fa-IR-DilaraNeural",
+    "Farid": "fa-IR-FaridNeural"
+  },
+  "Finnish": {
+    "Harri": "fi-FI-HarriNeural",
+    "Noora": "fi-FI-NooraNeural"
+  },
+  "Irish": {
+    "Colm": "ga-IE-ColmNeural",
+    "Orla": "ga-IE-OrlaNeural"
+  },
+  "Galician": {
+    "Roi": "gl-ES-RoiNeural",
+    "Sabela": "gl-ES-SabelaNeural"
+  },
+  "Gujarati": {
+    "Dhwani": "gu-IN-DhwaniNeural",
+    "Niranjan": "gu-IN-NiranjanNeural"
+  },
+  "Hindi": {
+    "Madhur": "hi-IN-MadhurNeural",
+    "Swara": "hi-IN-SwaraNeural"
+  },
+  "Croatian": {
+    "Gabrijela": "hr-HR-GabrijelaNeural",
+    "Srecko": "hr-HR-SreckoNeural"
+  },
+  "Hungarian": {
+    "Noemi": "hu-HU-NoemiNeural",
+    "Tamas": "hu-HU-TamasNeural"
+  },
+  "Icelandic": {
+    "Gudrun": "is-IS-GudrunNeural",
+    "Gunnar": "is-IS-GunnarNeural"
+  },
+  "Javanese": {
+    "Dimas": "jv-ID-DimasNeural",
+    "Siti": "jv-ID-SitiNeural"
+  },
+  "Georgian": {
+    "Eka": "ka-GE-EkaNeural",
+    "Giorgi": "ka-GE-GiorgiNeural"
+  },
+  "Kazakh": {
+    "Aigul": "kk-KZ-AigulNeural",
+    "Daulet": "kk-KZ-DauletNeural"
+  },
+  "Khmer": {
+    "Piseth": "km-KH-PisethNeural",
+    "Sreymom": "km-KH-SreymomNeural"
+  },
+  "Kannada": {
+    "Gagan": "kn-IN-GaganNeural",
+    "Sapna": "kn-IN-SapnaNeural"
+  },
+  "Lao": {
+    "Chanthavong": "lo-LA-ChanthavongNeural",
+    "Keomany": "lo-LA-KeomanyNeural"
+  },
+  "Lithuanian": {
+    "Leonas": "lt-LT-LeonasNeural",
+    "Ona": "lt-LT-OnaNeural"
+  },
+  "Latvian": {
+    "Everita": "lv-LV-EveritaNeural",
+    "Nils": "lv-LV-NilsNeural"
+  },
+  "Macedonian": {
+    "Aleksandar": "mk-MK-AleksandarNeural",
+    "Marija": "mk-MK-MarijaNeural"
+  },
+  "Malayalam": {
+    "Midhun": "ml-IN-MidhunNeural",
+    "Sobhana": "ml-IN-SobhanaNeural"
+  },
+  "Mongolian": {
+    "Bataa": "mn-MN-BataaNeural",
+    "Yesui": "mn-MN-YesuiNeural"
+  },
+  "Marathi": {
+    "Aarohi": "mr-IN-AarohiNeural",
+    "Manohar": "mr-IN-ManoharNeural"
+  },
+  "Maltese": {
+    "Grace": "mt-MT-GraceNeural",
+    "Joseph": "mt-MT-JosephNeural"
+  },
+  "Burmese": {
+    "Nilar": "my-MM-NilarNeural",
+    "Thiha": "my-MM-ThihaNeural"
+  },
+  "Nepali": {
+    "Hemkala": "ne-NP-HemkalaNeural",
+    "Sagar": "ne-NP-SagarNeural"
+  },
+  "Polish": {
+    "Marek": "pl-PL-MarekNeural",
+    "Zofia": "pl-PL-ZofiaNeural"
+  },
+  "Pashto": {
+    "Gul Nawaz": "ps-AF-GulNawazNeural",
+    "Latifa": "ps-AF-LatifaNeural"
+  },
+  "Romanian": {
+    "Alina": "ro-RO-AlinaNeural",
+    "Emil": "ro-RO-EmilNeural"
+  },
+  "Russian": {
+    "Svetlana": "ru-RU-SvetlanaNeural",
+    "Dmitry": "ru-RU-DmitryNeural"
+  },
+  "Sinhala": {
+    "Sameera": "si-LK-SameeraNeural",
+    "Thilini": "si-LK-ThiliniNeural"
+  },
+  "Slovak": {
+    "Lukas": "sk-SK-LukasNeural",
+    "Viktoria": "sk-SK-ViktoriaNeural"
+  },
+  "Slovenian": {
+    "Petra": "sl-SI-PetraNeural",
+    "Rok": "sl-SI-RokNeural"
+  },
+  "Somali": {
+    "Muuse": "so-SO-MuuseNeural",
+    "Ubax": "so-SO-UbaxNeural"
+  },
+  "Albanian": {
+    "Anila": "sq-AL-AnilaNeural",
+    "Ilir": "sq-AL-IlirNeural"
+  },
+  "Serbian": {
+    "Nicholas": "sr-RS-NicholasNeural",
+    "Sophie": "sr-RS-SophieNeural"
+  },
+  "Sundanese": {
+    "Jajang": "su-ID-JajangNeural",
+    "Tuti": "su-ID-TutiNeural"
+  },
+  "Swahili": {
+    "Rafiki": "sw-KE-RafikiNeural",
+    "Zuri": "sw-KE-ZuriNeural",
+    "Daudi": "sw-TZ-DaudiNeural",
+    "Rehema": "sw-TZ-RehemaNeural"
+    },
+  "Tamil": {
+    "Pallavi": "ta-IN-PallaviNeural",
+    "Valluvar": "ta-IN-ValluvarNeural",
+    "Kumar": "ta-LK-KumarNeural",
+    "Saranya": "ta-LK-SaranyaNeural",
+    "Kani": "ta-MY-KaniNeural",
+    "Surya": "ta-MY-SuryaNeural",
+    "Anbu": "ta-SG-AnbuNeural"
+  },
+  "Telugu": {
+    "Mohan": "te-IN-MohanNeural",
+    "Shruti": "te-IN-ShrutiNeural"
+  },
+  "Turkish": {
+    "Ahmet": "tr-TR-AhmetNeural",
+    "Emel": "tr-TR-EmelNeural"
+  },
+  "Ukrainian": {
+    "Ostap": "uk-UA-OstapNeural",
+    "Polina": "uk-UA-PolinaNeural"
+  },
+  "Urdu": {
+    "Gul": "ur-IN-GulNeural",
+    "Salman": "ur-IN-SalmanNeural",
+    "Asad": "ur-PK-AsadNeural",
+    "Uzma": "ur-PK-UzmaNeural"
+  },
+  "Uzbek": {
+    "Madina": "uz-UZ-MadinaNeural",
+    "Sardor": "uz-UZ-SardorNeural"
+  },
+  "Mandarin": {
+    "Xiaoxiao": "zh-CN-XiaoxiaoNeural",
+    "Yunyang": "zh-CN-YunyangNeural",
+    "Yunxi": "zh-CN-YunxiNeural",
+    "Xiaoyi": "zh-CN-XiaoyiNeural",
+    "Yunjian": "zh-CN-YunjianNeural",
+    "Yunxia": "zh-CN-YunxiaNeural",
+    "Xiaobei": "zh-CN-liaoning-XiaobeiNeural",
+    "Xiaoni": "zh-CN-shaanxi-XiaoniNeural",
+    "HiuMaan": "zh-HK-HiuMaanNeural",
+    "HiuGaai": "zh-HK-HiuGaaiNeural",
+    "WanLung": "zh-HK-WanLungNeural",
+    "HsiaoChen": "zh-TW-HsiaoChenNeural",
+    "HsiaoYu": "zh-TW-HsiaoYuNeural",
+    "YunJhe": "zh-TW-YunJheNeural"
+  },
+  "Zulu": {
+    "Thando": "zu-ZA-ThandoNeural",
+    "Themba": "zu-ZA-ThembaNeural"
+  }
+}
+
+client = Client("MohamedRashad/arabic-auto-tashkeel")
+
+async def text_to_speech_edge(text, language_code, speaker, tashkeel_checkbox=False):
+
+  # Remove diacritics from Arabic text then add tashkeel
+    if language_code == "Arabic" and tashkeel_checkbox:
+        text = client.predict(
+        		input_text=araby.strip_diacritics(text),
+        		api_name="/infer_shakkala"
+        )
+    
+    # Get the voice for the selected language and speaker
+    voice = language_dict[language_code][speaker]
+    communicate = edge_tts.Communicate(text, voice)
+    with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tmp_file:
+      tmp_path = tmp_file.name
+      await communicate.save(tmp_path)
+
+    return text, tmp_path
+
+
+def get_speakers(language):
+    print(language)
+    speakers = list(language_dict[language].keys())
+    return gr.Dropdown(choices=speakers, value=speakers[0], interactive=True), gr.Checkbox(visible=language == "Arabic", interactive=True)
+
+
+default_language = None
+default_speaker = None
+with gr.Blocks(title="Multilingual TTS") as demo:
+    gr.HTML("<center><h1>Multilingual TTS (Edge TTS)</h1></center>")
+    gr.HTML(f"<h2 style='color:Tomato;'> {len(language_dict)} languages supported</h3>")
+    gr.HTML(f"<p> {', '.join(language_dict.keys())} </h3>")
+    gr.Markdown("**Note:** A special feature is added for Arabic language only.")
+    with gr.Row():
+        with gr.Column():
+            input_text = gr.Textbox(lines=5, label="Input Text", placeholder="Enter text to convert to speech")
+            language = gr.Dropdown(
+                choices=list(language_dict.keys()), value=default_language, label="Languages", interactive=True
+            )
+            speaker = gr.Dropdown(choices=[], value=default_speaker, label="Speakers", interactive=False)
+            tashkeel_checkbox = gr.Checkbox(label="Tashkeel", value=False, visible=False, interactive=False)
+            run_btn = gr.Button(value="Generate Audio", variant="primary")
+
+        with gr.Column():
+            output_text = gr.Textbox(label="Output Text")
+            output_audio = gr.Audio(type="filepath", label="Audio Output")
+
+    language.change(get_speakers, inputs=[language], outputs=[speaker, tashkeel_checkbox])
+    run_btn.click(text_to_speech_edge, inputs=[input_text, language, speaker, tashkeel_checkbox], outputs=[output_text, output_audio])
+
+if __name__ == "__main__":
+    demo.queue().launch(share=False)