import spaces
import torch
import gradio as gr
from transformers import pipeline
from transformers.pipelines.audio_utils import ffmpeg_read
import tempfile
import os
import json

#===============
# Define main parameters
#===============

MODEL_NAME = "openai/whisper-large-v3-turbo"
BATCH_SIZE = 8
FILE_LIMIT_MB = 1000
YT_LENGTH_LIMIT_S = 3600  # limit to 1 hour YouTube files

device = 0 if torch.cuda.is_available() else "cpu"

pipe = pipeline(
    task="automatic-speech-recognition",
    model=MODEL_NAME,
    chunk_length_s=30,
    device=device,
)

#===============
# Main functions
#===============

@spaces.GPU
def transcribe(audio_file, task, language, keywords):
    if audio_file is None:
        raise gr.Error("No audio file submitted! Please upload or record an audio file before submitting your request.")

    result = pipe(audio_file, batch_size=BATCH_SIZE, generate_kwargs={"task": task, "language": language}, return_timestamps=True)
    text = result["text"]
    timestamps = result["chunks"]
    
    # First 200 characters for display
    preview_text = text[:200] + "..." if len(text) > 200 else text
    
    # Full transcription with timestamps in JSON
    full_transcription = {
        "text": text,
        "timestamps": timestamps
    }
    
    # Save the full transcription (with timestamps) as JSON
    json_file_path = tempfile.NamedTemporaryFile(delete=False, suffix=".json").name
    with open(json_file_path, "w") as json_file:
        json.dump(full_transcription, json_file)
    
    # Save the plain text transcription as TXT
    txt_file_path = tempfile.NamedTemporaryFile(delete=False, suffix=".txt").name
    with open(txt_file_path, "w") as txt_file:
        txt_file.write(text)
    
    return preview_text, json_file_path, txt_file_path

#===============
# Build the frontend
#===============

file_transcribe = gr.Interface(
    fn=transcribe,
    inputs=[
        gr.Audio(sources="upload", type="filepath", label="Audio file"),
        gr.Radio(["transcribe", "translate"], label="Task", value="transcribe"),
        gr.Dropdown(["spanish", "english"], label="Language", info="Will add more later!", value="spanish"),
        gr.Textbox(lines=10, label="Keywords"),
    ],
    outputs=[
        gr.Textbox(label="Preview (first 200 characters)"),
        gr.File(label="Download full transcription as JSON"),
        gr.File(label="Download transcription as TXT")
    ],
    title="Whisper Large V3: Transcribe Audio",
    description=(
        "Transcribe long-form audio inputs with the click of a button! Demo uses the"
        f" checkpoint [{MODEL_NAME}](https://huggingface.co./{MODEL_NAME}) and 🤗 Transformers to transcribe audio files"
        " of arbitrary length."
    ),
    allow_flagging="never",
)

#===============
# Launch
#===============

demo = gr.Blocks(theme=gr.themes.Ocean())

with demo:
    gr.TabbedInterface([file_transcribe], ["Audio file"])
    
demo.queue().launch(ssr_mode=False)