# -*- coding: utf-8 -*- from ttsmms import TTS import gradio as gr ISO_CODES = {'Tachelhit': 'shi', 'Tarifit (Latin script)': 'rif-script_latin', 'Tarifit (Arabic script)': 'rif-script_arabic', 'Taqbaylit': 'kab', 'Tamasheq': 'taq', 'Tamajaq, Tawallammat (Tifinagh script)': 'ttq-script_tifinagh' } mapping = {'ɣ': 'ġ', 'v': 'ġ', 'c': 'š', 'x': 'ḫ', 'T': 'ṭ', 'S': 'ṣ', 'D': 'ḍ', 'H': 'ḥ', 'Z': 'ẓ', '3': 'ε', '7': 'ḥ', '9': 'q', 'gh': 'ġ', 'kh': 'ḫ' } MODELS = {} def tts(text, variant): variant_code = ISO_CODES[variant] if variant_code not in MODELS: MODELS[variant_code] = TTS(variant_code) model = MODELS[variant_code] if variant_code == 'shi': for key, value in mapping.items(): text = text.replace(key, value) audio = model.synthesis(text) return (audio['sampling_rate'], audio['x']) examples = [["Riɣ ad cceɣ s ufus kḍuɣ s inxar", "Tachelhit"], ["arraw n lhem yukr aɣ ihdumn nɣ", "Tachelhit"], ["wa tamɣart ma d ukan teskart ?", "Tachelhit"], ["ar d iṭṭar unẓar, ffuɣn d igḍaḍ, mmɣin d ijjign", "Tachelhit"], ["Egg lxir di timura, ad tafed di tiwwura.", "Tarifit (Latin script)"], ["Aqemmum iqnen ur ṯ-ttidfen izan.", "Tarifit (Latin script)"]] description = "Text-to-speech for Tachelhit, Tarifit, Taqbaylit, Tamasheq and Tamajaq (Tawallammat)." iface = gr.Interface( fn=tts, inputs=[ gr.Textbox( label="Text", value="Text to synthesize." ), gr.Dropdown( label="Variant", choices=list(ISO_CODES.keys()), value="Tachelhit" ) ], outputs=gr.Audio(label="Output", type="numpy"), examples=examples, title="🗣️ Tamazight Text-to-Speech with MMS (Massively Multilingual Speech) 🗣️", description=description, allow_flagging="manual", flagging_options=['error', 'bad-quality', 'wrong-pronounciation'], ) iface.launch()