import torch

import gradio as gr
from transformers import pipeline
from transformers.pipelines.audio_utils import ffmpeg_read

from transcription import fast_transcription, speech_to_text
from audio import normalizeAudio, separateVoiceInstrumental, mp3_to_wav, stereo_to_mono, cutaudio, compose_audio
from audio import overlay_audios, compose_audio, total_duration, append_wav_files
from helpers import guardar_en_archivo


def transcribe(audiofile, model):

    audio_path = audiofile[0].name

    audio_normalized_path = normalizeAudio(audio_path, ".wav")

    novocal_path, vocal_path = separateVoiceInstrumental(audio_normalized_path)

    novocal_path = mp3_to_wav(novocal_path, "novocal")
    vocal_path = mp3_to_wav(vocal_path, "vocal")

    out = fast_transcription(vocal_path, model, "es")
    transcript = "\n".join(out)
    #Archivo
    nombre_archivo = guardar_en_archivo(out)

    return audio_path, audio_normalized_path, vocal_path, novocal_path, transcript, nombre_archivo


transcribeI = gr.Interface(
    fn=transcribe,
    inputs=[
        gr.File(label="Upload Files", file_count="multiple"),
        gr.Radio(["base", "small", "medium", "large-v2"], label="Models", value="large-v2"),
    ],
    outputs=[gr.Audio(type="filepath", label="original"),
             gr.Audio(type="filepath", label="normalized"),
             gr.Audio(type="filepath", label="vocal"),
             gr.Audio(type="filepath", label="no_vocal"),
             gr.TextArea(label="Transcription"),
             gr.File(label="Archivo generado")
        ],
    theme="huggingface",
    title="Transcripción",
    description=(
        "Sound extraction, processing, and dialogue transcription.\n"
        "Paste a link to a youtube video\n"
    ),
    allow_flagging="never",
    #examples=[[None, "COSER-4004-01-00_5m.wav", "large-v2"]]

)

demo = gr.Blocks()
with demo:
    gr.Markdown("# Dubbing")
    gr.TabbedInterface([diarizationI], ["Diarización"])

#demo.queue(concurrency_count=1).launch(enable_queue=True, auth=(os.environ['USER'], os.environ['PASSWORD']))
demo.launch(enable_queue=True)