app.py

import gradio as gr
from moviepy.editor import VideoFileClip
import speech_recognition as sr
from argostranslate import package, translate
import os
import tempfile

# ---------------------------
# Setup Argos Translate (offline)
# ---------------------------
def setup_translation():
    # Download English->Urdu model if not exists
    if not os.path.exists("en_ur.argosmodel"):
        import urllib.request
        url = "https://www.argosopentech.com/argospm/models/en_ur.argosmodel"
        urllib.request.urlretrieve(url, "en_ur.argosmodel")
        package.install_from_path("en_ur.argosmodel")
setup_translation()

# ---------------------------
# Extract audio from video
# ---------------------------
def extract_audio(video_path):
    clip = VideoFileClip(video_path)
    audio_path = tempfile.mktemp(suffix=".wav")
    clip.audio.write_audiofile(audio_path, fps=16000, codec="pcm_s16le")
    return audio_path

# ---------------------------
# Transcribe audio using CMU Sphinx
# ---------------------------
def transcribe_audio(audio_path):
    r = sr.Recognizer()
    with sr.AudioFile(audio_path) as source:
        audio = r.record(source)
    try:
        text = r.recognize_sphinx(audio)
        return text
    except sr.UnknownValueError:
        return "[Could not understand audio]"
    except sr.RequestError as e:
        return f"[Sphinx error: {e}]"

# ---------------------------
# Translate text using Argos Translate
# ---------------------------
def translate_text(text, target_lang):
    if target_lang == "original":
        return text
    from_lang = "en"
    to_lang = target_lang
    installed_languages = translate.get_installed_languages()
    from_lang_obj = next((l for l in installed_languages if l.code == from_lang), None)
    to_lang_obj = next((l for l in installed_languages if l.code == to_lang), None)
    if from_lang_obj and to_lang_obj:
        translated = from_lang_obj.get_translation(to_lang_obj).translate(text)
        return translated
    return text

# ---------------------------
# Gradio Interface
# ---------------------------
languages = {
    "original": "No Translation",
    "ur": "Urdu",
    "hi": "Hindi",
    "ps": "Pashto",
    "ar": "Arabic",
    "en": "English"
}

def process_video(video_file, lang):
    error_log = ""
    try:
        audio_path = extract_audio(video_file.name)
        error_log += "Audio extracted!\n"
    except Exception as e:
        return "", f"Audio extraction error: {e}"

    try:
        text = transcribe_audio(audio_path)
        error_log += f"Transcribed text length: {len(text)}\n"
    except Exception as e:
        return "", f"STT Error: {e}"

    try:
        translated = translate_text(text, lang)
        error_log += f"Translation done!\n"
    except Exception as e:
        translated = text
        error_log += f"Translation error: {e}\n"

    return translated, error_log

demo = gr.Interface(
    fn=process_video,
    inputs=[
        gr.Video(label="Upload Video"),
        gr.Dropdown(list(languages.keys()), value="original", label="Translate To")
    ],
    outputs=[
        gr.Textbox(label="Transcribed / Translated Text", interactive=False),
        gr.Textbox(label="Debug / Error Log", interactive=False)
    ],
    title="Offline Video Subtitle Generator",
    description="Upload a video → Extract audio → Generate subtitles → Optional translation → All offline, token-free"
)

demo.launch()