fairseq audio audio-to-audio speech-to-speech-translation

xm_transformer_600m-fr_en-multi_domain

W2V2-Transformer speech-to-text translation model from fairseq S2T (paper/code):

Usage

from fairseq.checkpoint_utils import load_model_ensemble_and_task_from_hf_hub
from fairseq.models.text_to_speech.hub_interface import S2THubInterface
from fairseq.models.text_to_speech.hub_interface import TTSHubInterface
import IPython.display as ipd
import torchaudio


models, cfg, task = load_model_ensemble_and_task_from_hf_hub(
    "facebook/xm_transformer_600m-fr_en-multi_domain",
    arg_overrides={"config_yaml": "config.yaml"},
)
model = models[0]
generator = task.build_generator(model, cfg)


# requires 16000Hz mono channel audio
audio, _ = torchaudio.load("/path/to/an/audio/file")

sample = S2THubInterface.get_model_input(task, audio)
text = S2THubInterface.get_prediction(task, model, generator, sample)

# speech synthesis
tts_models, tts_cfg, tts_task = load_model_ensemble_and_task_from_hf_hub(
  f"facebook/fastspeech2-en-ljspeech",
  arg_overrides={"vocoder": "griffin_lim", "fp16": False},
)
tts_model = tts_models[0]
TTSHubInterface.update_cfg_with_data_cfg(tts_cfg, tts_task.data_cfg)
tts_generator = tts_task.build_generator([tts_model], tts_cfg)

tts_sample = TTSHubInterface.get_model_input(tts_task, text)
wav, sr = TTSHubInterface.get_prediction(
    tts_task, tts_model, tts_generator, tts_sample
)

ipd.Audio(wav, rate=rate)

Citation

@inproceedings{li-etal-2021-multilingual,
    title = "Multilingual Speech Translation from Efficient Finetuning of Pretrained Models",
    author = "Li, Xian  and
      Wang, Changhan  and
      Tang, Yun  and
      Tran, Chau  and
      Tang, Yuqing  and
      Pino, Juan  and
      Baevski, Alexei  and
      Conneau, Alexis  and
      Auli, Michael",
    booktitle = "Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)",
    month = aug,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.acl-long.68",
    doi = "10.18653/v1/2021.acl-long.68",
    pages = "827--838",
}

@inproceedings{wang-etal-2020-fairseq,
    title = "Fairseq {S}2{T}: Fast Speech-to-Text Modeling with Fairseq",
    author = "Wang, Changhan  and
      Tang, Yun  and
      Ma, Xutai  and
      Wu, Anne  and
      Okhonko, Dmytro  and
      Pino, Juan",
    booktitle = "Proceedings of the 1st Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 10th International Joint Conference on Natural Language Processing: System Demonstrations",
    month = dec,
    year = "2020",
    address = "Suzhou, China",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2020.aacl-demo.6",
    pages = "33--39",
}

@inproceedings{wang-etal-2021-fairseq,
    title = "fairseq S{\^{}}2: A Scalable and Integrable Speech Synthesis Toolkit",
    author = "Wang, Changhan  and
      Hsu, Wei-Ning  and
      Adi, Yossi  and
      Polyak, Adam  and
      Lee, Ann  and
      Chen, Peng-Jen  and
      Gu, Jiatao  and
      Pino, Juan",
    booktitle = "Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing: System Demonstrations",
    month = nov,
    year = "2021",
    address = "Online and Punta Cana, Dominican Republic",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.emnlp-demo.17",
    doi = "10.18653/v1/2021.emnlp-demo.17",
    pages = "143--152",
}