app.py

# -*- coding: utf-8 -*-
from ttsmms import TTS
import gradio as gr

VARIANTS = ['shi', 'rif-script_latin', 'rif-script_arabic', 'kab', 'taq', 'ttq-script_tifinagh']
MODELS = {}

def tts(text, variant):
    if variant not in MODELS:
        MODELS[variant] = TTS(variant)
    model = MODELS[variant]
    audio = model.synthesis(text)
    return (audio['sampling_rate'], audio['x'])

examples = [["wala manis a-ttidun?", "shi"],
            ["ġ-iḍ-an ġ-ilul-umsiggel, illas lḥal s-umdlu isemmiḍn.", "shi"]]

iface = gr.Interface(
    fn=tts,
    inputs=[
        gr.inputs.Textbox(
            label="Text",
            default="Text to synthesize.",
        ),
        gr.inputs.Dropdown(label="Select a variant", choices=VARIANTS, default=VARIANTS[0])
    ],
    outputs=gr.outputs.Audio(label="Output", type="numpy"),
    examples=examples,
    title="🗣️ Tamazight Text to Speech with MMS 🗣️",
    allow_flagging="manual",
    flagging_options=['error', 'bad-quality', 'wrong-pronounciation'],
)
iface.launch()