models/f5tts_model.py

from f5_tts.api import F5TTS
import tempfile

from config.settings import (
    REFERENCE_AUDIO,
    REFERENCE_TEXT
)

print("Loading F5-TTS...")

model = F5TTS()

print("F5-TTS loaded")

def generate(sentence):
    
    output_wav = tempfile.mktemp(
        suffix = ".wav"
    )

    model.infer(
        ref_file = REFERENCE_AUDIO,
        ref_text = REFERENCE_TEXT,
        gen_text = sentence,
        file_wave = output_wav,
        remove_silence = False,
    )

    return output_wav