app.py

import torch
import torchaudio
from transformers import WhisperProcessor, WhisperForConditionalGeneration
from transformers import MarianMTModel, MarianTokenizer
import gradio as gr
from pydub import AudioSegment
import os

# Load Whisper model for transcription
whisper_model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-small")
whisper_processor = WhisperProcessor.from_pretrained("openai/whisper-small")
whisper_model.eval()
torch.set_grad_enabled(False)

# Load MarianMT model for translation (English → Spanish)
translation_model_name = "Helsinki-NLP/opus-mt-en-es"
translator = MarianMTModel.from_pretrained(translation_model_name)
tokenizer = MarianTokenizer.from_pretrained(translation_model_name)

def translate_text(text):
    inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True)
    translated = translator.generate(**inputs)
    return tokenizer.decode(translated[0], skip_special_tokens=True)

def transcribe_and_translate(filepath):
    if filepath is None or not os.path.exists(filepath):
        return "No audio file provided."

    audio = AudioSegment.from_file(filepath).set_channels(1)
    chunk_length_ms = 30 * 1000
    chunks = [audio[i:i + chunk_length_ms] for i in range(0, len(audio), chunk_length_ms)]

    full_transcript = ""

    for i, chunk in enumerate(chunks):
        chunk_path = f"chunk_{i}.wav"
        chunk.export(chunk_path, format="wav")
        waveform, sr = torchaudio.load(chunk_path)
        os.remove(chunk_path)

        if sr != 16000:
            waveform = torchaudio.transforms.Resample(orig_freq=sr, new_freq=16000)(waveform)

        waveform = waveform.mean(dim=0)
        inputs = whisper_processor(waveform, sampling_rate=16000, return_tensors="pt")
        predicted_ids = whisper_model.generate(inputs["input_features"])
        transcription = whisper_processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]

        full_transcript += transcription.strip() + " "

    translated_text = translate_text(full_transcript.strip())
    return translated_text

mic_ui = gr.Interface(
    fn=transcribe_and_translate,
    inputs=gr.Audio(sources="microphone", type="filepath"),
    outputs=gr.Textbox(label="Translation (English Speech → Spanish Text)"),
)

file_ui = gr.Interface(
    fn=transcribe_and_translate,
    inputs=gr.Audio(sources="upload", type="filepath"),
    outputs=gr.Textbox(label="Translation (English Speech → Spanish Text)"),
)

app = gr.TabbedInterface([mic_ui, file_ui], ["Microphone Input", "Upload File"])
app.launch()