Usage

from fairseq.checkpoint_utils import load_model_ensemble_and_task_from_hf_hub
from fairseq.models.text_to_speech.hub_interface import TTSHubInterface
import IPython.display as ipd

models, cfg, task = load_model_ensemble_and_task_from_hf_hub(
    "amaai-lab/DisfluencySpeech_BenchmarkA",
    arg_overrides={"vocoder": "hifigan", "fp16": False, "spec-bwd-max-iter": 32}
)
model = models[0]
TTSHubInterface.update_cfg_with_data_cfg(cfg, task.data_cfg)
generator = task.build_generator(models, cfg)

text = "Well, that's really funny, isn't it? What a strange world we live in."

sample = TTSHubInterface.get_model_input(task, text)
sample['net_input']['src_tokens'] = sample['net_input']['src_tokens'].cuda()
sample['net_input']['src_lengths'] = sample['net_input']['src_lengths'].cuda()
wav, rate = TTSHubInterface.get_prediction(task, model.cuda(), generator, sample)

ipd.Audio(wav.cpu(), rate=rate)