Create app.py

app.py

@@ -0,0 +1,109 @@
+import gradio as gr
+import os
+import torch
+import torchaudio
+from transformers import (
+    WhisperProcessor, WhisperForConditionalGeneration,
+    SpeechT5Processor, SpeechT5ForTextToSpeech,
+    MarianMTModel, MarianTokenizer
+)
+import ffmpeg
+import soundfile as sf
+
+# Configurações
+UPLOAD_FOLDER = "uploads"
+OUTPUT_FOLDER = "outputs"
+
+if not os.path.exists(UPLOAD_FOLDER):
+    os.makedirs(UPLOAD_FOLDER)
+if not os.path.exists(OUTPUT_FOLDER):
+    os.makedirs(OUTPUT_FOLDER)
+
+# Verificar se a GPU está disponível
+device = "cuda" if torch.cuda.is_available() else "cpu"
+print(f"Usando dispositivo: {device}")
+
+# Inicializar modelos
+whisper_processor = WhisperProcessor.from_pretrained("openai/whisper-small")
+whisper_model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-small").to(device)
+
+tts_processor = SpeechT5Processor.from_pretrained("microsoft/speecht5_tts")
+tts_model = SpeechT5ForTextToSpeech.from_pretrained("microsoft/speecht5_tts").to(device)
+
+translation_model_name = "Helsinki-NLP/opus-mt-en-pt"
+translation_tokenizer = MarianTokenizer.from_pretrained(translation_model_name)
+translation_model = MarianMTModel.from_pretrained(translation_model_name).to(device)
+
+# Funções auxiliares
+def transcribe_audio(audio_path):
+    waveform, sample_rate = torchaudio.load(audio_path)
+    waveform = waveform.to(device)
+    inputs = whisper_processor(waveform.squeeze().cpu().numpy(), sampling_rate=sample_rate, return_tensors="pt").to(device)
+    with torch.no_grad():
+        predicted_ids = whisper_model.generate(inputs.input_features)
+    transcription = whisper_processor.batch_decode(predicted_ids, skip_special_tokens=True)
+    return transcription[0]
+
+def synthesize_speech(text, output_path):
+    inputs = tts_processor(text, return_tensors="pt").to(device)
+    with torch.no_grad():
+        speech = tts_model.generate_speech(inputs["input_ids"], tts_model.speaker_embeddings)
+    sf.write(output_path, speech.cpu().numpy(), samplerate=22050)
+
+def translate_text(text, target_language="pt"):
+    inputs = translation_tokenizer(text, return_tensors="pt", truncation=True).to(device)
+    with torch.no_grad():
+        translated_ids = translation_model.generate(**inputs)
+    translated_text = translation_tokenizer.decode(translated_ids[0], skip_special_tokens=True)
+    return translated_text
+
+def extract_audio(video_path, audio_path):
+    ffmpeg.input(video_path).output(audio_path, ac=1, ar="16000").run(overwrite_output=True)
+
+def replace_audio_in_video(video_path, audio_path, output_path):
+    video = ffmpeg.input(video_path)
+    audio = ffmpeg.input(audio_path)
+    ffmpeg.output(video.video, audio.audio, output_path, vcodec="copy", acodec="aac").run(overwrite_output=True)
+
+# Função principal para Gradio
+def translate_video(video):
+    video_path = os.path.join(UPLOAD_FOLDER, "input_video.mp4")
+    with open(video_path, "wb") as f:
+        f.write(video)
+
+    try:
+        # Extrair áudio do vídeo
+        audio_path = os.path.join(UPLOAD_FOLDER, "audio.wav")
+        extract_audio(video_path, audio_path)
+
+        # Transcrever áudio
+        transcribed_text = transcribe_audio(audio_path)
+        print("Texto transcrito:", transcribed_text)
+
+        # Traduzir texto
+        translated_text = translate_text(transcribed_text, target_language="pt")
+        print("Texto traduzido:", translated_text)
+
+        # Sintetizar áudio em português
+        synthesized_audio_path = os.path.join(UPLOAD_FOLDER, "synthesized_audio.wav")
+        synthesize_speech(translated_text, synthesized_audio_path)
+
+        # Substituir áudio no vídeo
+        output_video_path = os.path.join(OUTPUT_FOLDER, "translated_video.mp4")
+        replace_audio_in_video(video_path, synthesized_audio_path, output_video_path)
+
+        return output_video_path
+
+    except Exception as e:
+        return f"Erro: {str(e)}"
+
+# Interface Gradio
+iface = gr.Interface(
+    fn=translate_video,
+    inputs=gr.Video(),
+    outputs=gr.Video(),
+    title="Tradutor de Vídeo",
+    description="Carregue um vídeo em qualquer idioma e ele será traduzido para português."
+)
+
+iface.launch()