---
library_name: fairseq
task: audio-to-audio
tags:
- fairseq
- audio
- audio-to-audio
- speech-to-speech-translation

datasets:
- mtedx
- covost2
- europarl_st
- voxpopuli
widget:
- example_title: Common Voice sample 1
  src: https://huggingface.co/facebook/xm_transformer_600m-es_en-multi_domain/resolve/main/common_voice_es_19966634.flac
---
# xm_transformer_600m-es_en-multi_domain

[W2V2-Transformer](https://aclanthology.org/2021.acl-long.68/) speech-to-text translation model from fairseq S2T ([paper](https://arxiv.org/abs/2010.05171)/[code](https://github.com/pytorch/fairseq/tree/main/examples/speech_to_text)):
- Spanish-English
- Trained on mTEDx, CoVoST 2, EuroParl-ST, VoxPopuli, Multilingual LibriSpeech, Common Voice v7 and CCMatrix
- Speech synthesis with [facebook/fastspeech2-en-ljspeech](https://huggingface.co/facebook/fastspeech2-en-ljspeech)

## Usage
```python
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-es_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)
```