app.py

import gradio as gr
import whisper
from googletrans import Translator
import srt
import datetime
import tempfile
import os

# Load model once
model = whisper.load_model("base")
translator = Translator()

def transcribe(audio):
    try:
        # Check if audio file exists
        if audio is None:
            return "❌ No audio uploaded!", None

        # Transcribe audio
        result = model.transcribe(audio)
        segments = result.get("segments", [])
        subtitles = []
        full_text = ""

        for i, seg in enumerate(segments):
            start = datetime.timedelta(seconds=float(seg["start"]))
            end = datetime.timedelta(seconds=float(seg["end"]))
            text = seg["text"].strip()

            # Translate Hindi → Hinglish (pronunciation)
            translated = translator.translate(text, src='hi', dest='en').pronunciation
            if not translated:
                translated = translator.translate(text, src='hi', dest='en').text

            subtitles.append(srt.Subtitle(index=i+1, start=start, end=end, content=translated))
            full_text += translated + " "

        # Create SRT file
        srt_content = srt.compose(subtitles)

        # Temporary file save
        srt_path = os.path.join(tempfile.gettempdir(), "output.srt")
        with open(srt_path, "w", encoding="utf-8") as f:
            f.write(srt_content)

        return full_text.strip(), srt_path

    except Exception as e:
        return f"❌ Runtime Error: {str(e)}", None


iface = gr.Interface(
    fn=transcribe,
    inputs=gr.Audio(source="upload", type="filepath", label="🎤 Upload Hindi Audio"),
    outputs=[
        gr.Textbox(label="📝 Hinglish Subtitles"),
        gr.File(label="⬇️ Download SRT File")
    ],
    title="Hinglish Subtitle Generator 🇮🇳",
    description="Upload Hindi speech audio — get Hinglish subtitles (.srt) instantly!"
)

iface.launch()