def test_no_espeak_phonemes():
# prepare the config
config = VitsConfig(
batch_size=2,
eval_batch_size=2,
num_loader_workers=0,
num_eval_loader_workers=0,
text_cleaner="english_cleaners",
use_phonemes=True,
use_espeak_phonemes=False,
phoneme_language="en-us",
phoneme_cache_path="tests/data/ljspeech/phoneme_cache/",
run_eval=True,
test_delay_epochs=-1,
epochs=1,
print_step=1,
print_eval=True,
datasets=[dataset_config_en, dataset_config_pt],
)
config.save_json(config_path)
# run test
run_cli(
f'CUDA_VISIBLE_DEVICES="" python TTS/bin/find_unique_phonemes.py --config_path "{config_path}"'
)
def test_synthesize():
"""Test synthesize.py with diffent arguments."""
output_path = os.path.join(get_tests_output_path(), "output.wav")
run_cli("tts --list_models")
# single speaker model
run_cli(f'tts --text "This is an example." --out_path "{output_path}"')
def run_test_train():
command = (
f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_encoder.py --config_path {config_path} "
f"--coqpit.output_path {output_path} "
"--coqpit.datasets.0.name ljspeech "
"--coqpit.datasets.0.meta_file_train metadata.csv "
"--coqpit.datasets.0.meta_file_val metadata.csv "
"--coqpit.datasets.0.path tests/data/ljspeech ")
run_cli(command)
def test_run_all_models():
"""Check if all the models are downloadable and tts models run correctly."""
print(" > Run synthesizer with all the models.")
download_dir = get_user_data_dir("tts")
output_path = os.path.join(get_tests_output_path(), "output.wav")
manager = ModelManager(output_prefix=get_tests_output_path())
model_names = manager.list_models()
for model_name in model_names:
print(f"\n > Run - {model_name}")
model_path, _, _ = manager.download_model(model_name)
if "tts_models" in model_name:
local_download_dir = os.path.dirname(model_path)
# download and run the model
speaker_files = glob.glob(local_download_dir + "/speaker*")
language_files = glob.glob(local_download_dir + "/language*")
language_id = ""
if len(speaker_files) > 0:
# multi-speaker model
if "speaker_ids" in speaker_files[0]:
speaker_manager = SpeakerManager(
speaker_id_file_path=speaker_files[0])
elif "speakers" in speaker_files[0]:
speaker_manager = SpeakerManager(
d_vectors_file_path=speaker_files[0])
# multi-lingual model - Assuming multi-lingual models are also multi-speaker
if len(language_files
) > 0 and "language_ids" in language_files[0]:
language_manager = LanguageManager(
language_ids_file_path=language_files[0])
language_id = language_manager.language_names[0]
speaker_id = list(speaker_manager.ids.keys())[0]
run_cli(
f"tts --model_name {model_name} "
f'--text "This is an example." --out_path "{output_path}" --speaker_idx "{speaker_id}" --language_idx "{language_id}" '
)
else:
# single-speaker model
run_cli(
f"tts --model_name {model_name} "
f'--text "This is an example." --out_path "{output_path}"')
# remove downloaded models
shutil.rmtree(download_dir)
else:
# only download the model
manager.download_model(model_name)
print(f" | > OK: {model_name}")
folders = glob.glob(os.path.join(manager.output_prefix, "*"))
assert len(folders) == len(model_names)
shutil.rmtree(manager.output_prefix)
def test_voice_conversion():
print(" > Run voice conversion inference using YourTTS model.")
model_name = "tts_models/multilingual/multi-dataset/your_tts"
language_id = "en"
speaker_wav = os.path.join(get_tests_data_path(), "ljspeech", "wavs",
"LJ001-0001.wav")
reference_wav = os.path.join(get_tests_data_path(), "ljspeech", "wavs",
"LJ001-0032.wav")
output_path = os.path.join(get_tests_output_path(), "output.wav")
run_cli(
f"tts --model_name {model_name}"
f" --out_path {output_path} --speaker_wav {speaker_wav} --reference_wav {reference_wav} --language_idx {language_id} "
)
def test_Tacotron():
# set paths
config_path = os.path.join(get_tests_input_path(), "test_tacotron_config.json")
checkpoint_path = os.path.join(get_tests_output_path(), "checkpoint_test.pth.tar")
output_path = os.path.join(get_tests_output_path(), "output_extract_tts_spectrograms/")
# load config
c = load_config(config_path)
# create model
num_chars = len(phonemes if c.use_phonemes else symbols)
model = setup_model(num_chars, 1, c, speaker_embedding_dim=None)
# save model
torch.save({"model": model.state_dict()}, checkpoint_path)
# run test
run_cli(
f'CUDA_VISIBLE_DEVICES="" python TTS/bin/extract_tts_spectrograms.py --config_path "{config_path}" --checkpoint_path "{checkpoint_path}" --output_path "{output_path}"'
)
run_cli(f'rm -rf "{output_path}" "{checkpoint_path}"')
def test_synthesize():
"""Test synthesize.py with diffent arguments."""
output_path = os.path.join(get_tests_output_path(), "output.wav")
run_cli("tts --list_models")
# single speaker model
run_cli(f'tts --text "This is an example." --out_path "{output_path}"')
run_cli("tts --model_name tts_models/en/ljspeech/glow-tts "
f'--text "This is an example." --out_path "{output_path}"')
run_cli("tts --model_name tts_models/en/ljspeech/glow-tts "
"--vocoder_name vocoder_models/en/ljspeech/multiband-melgan "
f'--text "This is an example." --out_path "{output_path}"')
def test_Tacotron():
# set paths
config_path = os.path.join(get_tests_input_path(),
"test_tacotron_config.json")
checkpoint_path = os.path.join(get_tests_output_path(),
"checkpoint_test.pth")
output_path = os.path.join(get_tests_output_path(),
"output_extract_tts_spectrograms/")
# load config
c = load_config(config_path)
# create model
model = setup_model(c)
# save model
torch.save({"model": model.state_dict()}, checkpoint_path)
# run test
run_cli(
f'CUDA_VISIBLE_DEVICES="" python TTS/bin/extract_tts_spectrograms.py --config_path "{config_path}" --checkpoint_path "{checkpoint_path}" --output_path "{output_path}"'
)
run_cli(f'rm -rf "{output_path}" "{checkpoint_path}"')
def test():
# set paths
wav_path = os.path.join(get_tests_input_path(),
"../data/ljspeech/wavs")
output_path = os.path.join(get_tests_output_path(),
"output_wavs_removed_silence/")
output_resample_path = os.path.join(get_tests_output_path(),
"output_ljspeech_16khz/")
# resample audios
run_cli(
f'CUDA_VISIBLE_DEVICES="" python TTS/bin/resample.py --input_dir "{wav_path}" --output_dir "{output_resample_path}" --output_sr 16000'
)
# run test
run_cli(
f'CUDA_VISIBLE_DEVICES="" python TTS/bin/remove_silence_using_vad.py --input_dir "{output_resample_path}" --output_dir "{output_path}"'
)
run_cli(f'rm -rf "{output_resample_path}"')
run_cli(f'rm -rf "{output_path}"')
num_loader_workers=0,
num_val_loader_workers=0,
run_eval=True,
test_delay_epochs=-1,
epochs=1,
seq_len=8192,
eval_split_size=1,
print_step=1,
print_eval=True,
data_path="tests/data/ljspeech",
output_path=output_path,
test_noise_schedule={"min_val": 1e-6, "max_val": 1e-2, "num_steps": 2},
)
config.audio.do_trim_silence = True
config.audio.trim_db = 60
config.save_json(config_path)
# train the model for one epoch
command_train = (
f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_wavegrad.py --config_path {config_path} "
)
run_cli(command_train)
# Find latest folder
continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
# restore the model and continue training for one more epoch
command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_wavegrad.py --continue_path {continue_path} "
run_cli(command_train)
shutil.rmtree(continue_path)
config.use_speaker_embedding = True
config.model_args.use_speaker_embedding = True
config.audio.trim_db = 60
config.save_json(config_path)
# train the model for one epoch
command_train = (
f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --config_path {config_path} "
f"--coqpit.output_path {output_path} "
"--coqpit.datasets.0.name ljspeech_test "
"--coqpit.datasets.0.meta_file_train metadata.csv "
"--coqpit.datasets.0.meta_file_val metadata.csv "
"--coqpit.datasets.0.path tests/data/ljspeech "
"--coqpit.datasets.0.meta_file_attn_mask tests/data/ljspeech/metadata_attn_mask.txt "
"--coqpit.test_delay_epochs 0")
run_cli(command_train)
# Find latest folder
continue_path = max(glob.glob(os.path.join(output_path, "*/")),
key=os.path.getmtime)
# Inference using TTS API
continue_config_path = os.path.join(continue_path, "config.json")
continue_restore_path, _ = get_last_checkpoint(continue_path)
out_wav_path = os.path.join(get_tests_output_path(), "output.wav")
speaker_id = "ljspeech-1"
continue_speakers_path = os.path.join(continue_path, "speakers.json")
# Check integrity of the config
with open(continue_config_path, "r", encoding="utf-8") as f:
config_loaded = json.load(f)
