app.py

import gradio as gr

import os

from gradio_client import Client

from dotenv import load_dotenv
load_dotenv()


HF_TOKEN = os.environ['hf_token']

client = Client("Ghana-NLP/Naija-Senegal-TTS",hf_token=HF_TOKEN)

def synthesize(text, lang, speaker):
    audio = result = client.predict(
        text=text,
		lang=lang,
		speaker=speaker,
		api_name="/predict")
    return audio


title = "Khaya AI: Speech Synthesis for Naija, Senegal & Northern Ghana"

description = """
<b>How to use:</b> Enter some text and choose a language and speaker.
"""

examples = [
    ["Ubangiji Ubangiji ya albarkaci ƙarshen Ayuba fiye da farkonsa. Yana da tumaki guda dubu goma sha huɗu raƙuma guda dubu shida shanun noma guda dubu da jakuna guda dubu.", "Hausa","Male (Low)"],
    ["Wata rana da yake cikin koyarwa waɗansu Farisiyawa da malaman dokoki waɗanda suka zo daga kowane ƙauye a Galili Yahudiya da Urushalima kuwa suna zaune. Ikon Ubangiji yana nan don yă warkar da masu ciwo.", "Hausa","Female"],
    ["Ọ dịghị ihe nwere ike ime ka i si nʼọnọdụ ị nọ ugbu a biliekwa. Nʼihi na ọnya dị gị nʼahụ ga-egbu gị. Ndị niile nụrụ ihe dakwasịrị gị ga-eji ọṅụ kụọ aka ha, nʼihi na onye ka mmadụ ahụ bụ nke na-anụrụghị ụfụ nke enweghị obi ebere gị?", "Igbo","Male (Low)"],
    ["Ịtụ egwu Onyenwe anyị bụ ịkpọ ihe ọjọọ asị. Akpọrọ m nganga na mpako, omume ọjọọ nakwa okwu gbagọrọ agbagọ asị.", "Igbo","Female"],
    ["Teewul maa ngi leen di wax ne, Iliyas ñëw na te xàmmiwuñu ko, waaye def nañu ko la leen neex. Noonu it dañoo nar a sonal léegi Doomu nit ki.»", "Wolof","Male (Low)"],
    ["Bind naa leen, du ndax xamuleen dëgg, waaye ndax xam ngeen ko, xam it ne wenn fen du jóge ci dëgg.", "Wolof","Female"],
    ["Bí wọ́n ti bẹ̀rẹ̀ sí kọrin àti ìyìn, Olúwa rán ogun ẹ̀yìn sí àwọn ọkùnrin Ammoni àti Moabu àti òkè Seiri tí ó ń gbógun ti Juda, wọ́n sì kọlù wọ́n.", "Yoruba","Male (Low)"],
    ["Ṣùgbọ́n lẹ́yìn ìgbà tí wọ́n jẹ́ wọ́n tan, kò sì ẹni tí ó le mọ̀ pé wọ́n jẹ ohunkóhun, nítorí wọn kò sanra sí i, wọn sì bùrẹ́wà síbẹ̀. Nígbà náà ni mo tají.", "Yoruba","Female"],
    ["Dat nite di warda kon wosh all dia sore den en and en family kon baptize.", "Pidgin","Male (High)"]
    ["And wen dem divide evritin wey get life for di sea into three one part die finish. Den wen dem divide all di ship for di world into three one part burn finish too.", "Pidgin","Female"]
    ["Si dans une tribu, une femme reçoit des terres en héritage, elle doit se marier avec un homme appartenant à la tribu de son père. Ainsi, chaque tribu israélite gardera les terres reçues de ses ancêtres.", "French","Male (Low)"]
    ["And this model speaks English in a West African accent too, if you can believe that.", "English","Male (Low)"],
]

gr.Interface(
    fn=synthesize,
    inputs=[
        gr.Text(label="Input Text"),
        gr.Dropdown(
            label="Select Language",
            choices=["Hausa","Igbo","Pidgin","Wolof","Yoruba","English","French"],
            value="Pidgin",
        ),
        gr.Radio(label="Speaker", choices=[
            "Female",
            "Male (Low)",
            "Male (High)"
        ],
        value="Male (Low)"),
    ],
    outputs=[
        gr.Audio(label="Generated Speech", type="numpy"),
    ],
    title=title,
    description=description,
    examples=examples,
).launch()