facebook

/

fastspeech2-en-ljspeech

like 268

Follow

AI at Meta 4.55k

Use the model=models it worked for me

I had multiple issues on OSX. This code worked:

from fairseq.checkpoint_utils import load_model_ensemble_and_task_from_hf_hub
from fairseq.models.text_to_speech.hub_interface import TTSHubInterface
import soundfile as sf

models, cfg, task = load_model_ensemble_and_task_from_hf_hub(
    "facebook/fastspeech2-en-ljspeech",
    arg_overrides={"vocoder": "hifigan", "fp16": False}
)

TTSHubInterface.update_cfg_with_data_cfg(cfg, task.data_cfg)
generator = task.build_generator(models, cfg)

text = "Hello, this is a test run."

sample = TTSHubInterface.get_model_input(task, text)
wav, rate = TTSHubInterface.get_prediction(task, models[0], generator, sample)

sf.write("speech.wav", wav, samplerate=rate)

This finally worked for me in colab
Started a fresh runtime then,

!pip install fairseq
!pip install huggingface_hub
!pip install g2p_en

from fairseq.checkpoint_utils import load_model_ensemble_and_task_from_hf_hub
from fairseq.models.text_to_speech.hub_interface import TTSHubInterface
import soundfile as sf

models, cfg, task = load_model_ensemble_and_task_from_hf_hub(
   "facebook/fastspeech2-en-ljspeech",
   arg_overrides={"vocoder": "hifigan", "fp16": False}
)

TTSHubInterface.update_cfg_with_data_cfg(cfg, task.data_cfg)
generator = task.build_generator(models, cfg)

text = "Hello, this is a test run."

sample = TTSHubInterface.get_model_input(task, text)
wav, rate = TTSHubInterface.get_prediction(task, models[0], generator, sample)

sf.write("speech.wav", wav, samplerate=rate)

Finally right.