app.py

import streamlit as st
from transformers import WhisperProcessor, WhisperForConditionalGeneration, RagTokenizer, RagRetriever, RagSequenceForGeneration
import torch
import soundfile as sf
import librosa
from moviepy.editor import VideoFileClip
import os
import tempfile
import logging

# Set up logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

# Load Whisper base model and processor
whisper_model_name = "openai/whisper-base"
whisper_processor = WhisperProcessor.from_pretrained(whisper_model_name)
whisper_model = WhisperForConditionalGeneration.from_pretrained(whisper_model_name)

# Load RAG sequence model and tokenizer
rag_model_name = "facebook/rag-sequence-nq"
rag_tokenizer = RagTokenizer.from_pretrained(rag_model_name)

# Try to load RagRetriever with trust_remote_code=True
try:
    rag_retriever = RagRetriever.from_pretrained(
        rag_model_name,
        index_name="exact",
        use_dummy_dataset=True,
        dataset_path=local_dataset_path,
        trust_remote_code=True
    )
    logger.info("Successfully loaded RagRetriever with trust_remote_code=True")
except ValueError as e:
    logger.error(f"Error loading RagRetriever: {e}")
    rag_retriever = None

if rag_retriever is not None:
    rag_model = RagSequenceForGeneration.from_pretrained(rag_model_name, retriever=rag_retriever)
else:
    logger.error("RagRetriever is not available, unable to proceed with loading RAG model.")
    st.error("RagRetriever is not available, unable to proceed with loading RAG model.")

def transcribe_audio(audio_path, language="ru"):
    speech, rate = librosa.load(audio_path, sr=16000)
    inputs = whisper_processor(speech, return_tensors="pt", sampling_rate=16000)
    input_features = whisper_processor.feature_extractor(speech, return_tensors="pt", sampling_rate=16000).input_features
    predicted_ids = whisper_model.generate(input_features, forced_decoder_ids=whisper_processor.get_decoder_prompt_ids(language=language, task="translate"))
    transcription = whisper_processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
    return transcription

def translate_and_summarize(text):
    if rag_retriever is None:
        return ["Translation and summarization feature is not available due to RAG retriever loading issue."]
    inputs = rag_tokenizer(text, return_tensors="pt")
    input_ids = inputs["input_ids"]
    attention_mask = inputs["attention_mask"]
    outputs = rag_model.generate(input_ids=input_ids, attention_mask=attention_mask)
    return rag_tokenizer.batch_decode(outputs, skip_special_tokens=True)

def extract_audio_from_video(video_path, output_audio_path):
    video_clip = VideoFileClip(video_path)
    audio_clip = video_clip.audio
    if audio_clip is not None:
        audio_clip.write_audiofile(output_audio_path)
        return output_audio_path
    else:
        return None

st.title("Audio and Video Transcription & Summarization")

# Audio Upload Section
st.header("Upload an Audio File")
audio_file = st.file_uploader("Choose an audio file...", type=["wav", "mp3", "m4a"])

if audio_file is not None:
    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
        tmp_file.write(audio_file.getbuffer())
        audio_path = tmp_file.name
    
    st.audio(audio_file)
    st.write("Transcribing audio...")
    try:
        transcription = transcribe_audio(audio_path)
        st.write("Transcription:", transcription)
        
        st.write("Translating and summarizing...")
        summary = translate_and_summarize(transcription)
        st.write("Translated Summary:", summary)
    except Exception as e:
        st.error(f"An error occurred: {e}")

# Video Upload Section
st.header("Upload a Video File")
video_file = st.file_uploader("Choose a video file...", type=["mp4", "mkv", "avi", "mov"])

if video_file is not None:
    with tempfile.NamedTemporaryFile(delete=False, suffix=".mp4") as tmp_file:
        tmp_file.write(video_file.getbuffer())
        video_path = tmp_file.name
    
    st.video(video_file)
    st.write("Extracting audio from video...")
    audio_path = extract_audio_from_video(video_path, tempfile.NamedTemporaryFile(delete=False, suffix=".wav").name)
    
    if audio_path is not None:
        st.write("Transcribing audio...")
        try:
            transcription = transcribe_audio(audio_path)
            st.write("Transcription:", transcription)
            
            st.write("Translating and summarizing...")
            summary = translate_and_summarize(transcription)
            st.write("Translated Summary:", summary)
        except Exception as e:
            st.error(f"An error occurred: {e}")
    else:
        st.write("No audio track found in the video file.")