def test_hifigan_synthesis():
hifigan_model_path = os.path.join("test_samples", "hifigan.pt")
hifigan_config_path = os.path.join("test_samples", "config.json")
audio_path = "synthesized_audio.wav"
transcription_model = Silero()
hifigan = Hifigan(hifigan_model_path, hifigan_config_path)
text = "the monkeys live"
synthesize(
model=FakeModelForSynthesis(),
text=text,
graph_path=None,
audio_path=audio_path,
vocoder=hifigan,
)
assert os.path.isfile(audio_path)
assert similarity(
text, transcription_model.transcribe(audio_path)) > MIN_SYNTHESIS_SCORE
os.remove(audio_path)
def test_silero():
transcription_model = Silero()
audio_path = os.path.join("test_samples", "audio.wav")
transcription = transcription_model.transcribe(audio_path)
assert similarity(TEXT, transcription) > MIN_SYNTHESIS_SCORE