import gradio as gr
import librosa
from transformers import AutoFeatureExtractor, pipeline


def load_and_fix_data(input_file, model_sampling_rate):
    speech, sample_rate = librosa.load(input_file)
    if len(speech.shape) > 1:
        speech = speech[:, 0] + speech[:, 1]
    if sample_rate != model_sampling_rate:
        speech = librosa.resample(speech, sample_rate, model_sampling_rate)
    return speech


feature_extractor = AutoFeatureExtractor.from_pretrained(
    "anuragshas/wav2vec2-xls-r-1b-hi-with-lm"
)
sampling_rate = feature_extractor.sampling_rate

asr = pipeline("automatic-speech-recognition", model="anuragshas/wav2vec2-xls-r-1b-hi-with-lm")



model = AutoModelForSeq2SeqLM.from_pretrained('hackathon-pln-es/t5-small-spanish-nahuatl')
tokenizer = AutoTokenizer.from_pretrained('hackathon-pln-es/t5-small-spanish-nahuatl')


def predict_and_ctc_lm_decode(input_file):
    speech = load_and_fix_data(input_file, sampling_rate)
    transcribed_text = asr(speech, chunk_length_s=5, stride_length_s=1)
    transcribed_text = transcribed_text["text"]
    input_ids = tokenizer('translate Spanish to Nahuatl: ' + transcribed_text, return_tensors='pt').input_ids
    outputs = model.generate(input_ids, max_length=512)
    outputs = tokenizer.batch_decode(outputs, skip_special_tokens=True)[0]
    return outputs


gr.Interface(
    predict_and_ctc_lm_decode,
    inputs=[
        gr.inputs.Audio(source="microphone", type="filepath", label="Record your audio")
    ],
    outputs=[gr.outputs.Textbox()],
    examples=[["example1.wav"]],
    title="Spanish-Audio-Transcriptions-to-Nahuatl-Translation",
    article="<p><center><img src='........e'></center></p>",
    layout="horizontal",
    theme="huggingface",
).launch(enable_queue=True, cache_examples=True)