Spaces:
Running
Running
| import gradio as gr | |
| import os | |
| import subprocess | |
| from gtts import gTTS | |
| from pydub import AudioSegment | |
| from transformers import GPT2LMHeadModel, GPT2Tokenizer, pipeline | |
| from PIL import Image | |
| generator = pipeline('text-generation', model="checkpoints") | |
| tokenizer = GPT2Tokenizer.from_pretrained('checkpoints') | |
| os.environ["TOKENIZERS_PARALLELISM"] = "true" | |
| def generate_output(name, date_of_birth, progress=gr.Progress()): | |
| prompt = f"Bienvenido {name}:" | |
| input_tokens = tokenizer.encode(prompt, add_special_tokens=True) | |
| input_text = tokenizer.decode(input_tokens) | |
| gpt2_output = generator(input_text, max_length=60, do_sample=True, temperature=0.6) | |
| generated_text = gpt2_output[0]['generated_text'] | |
| generated_text = generated_text.replace(input_text, "").strip() | |
| if len(gpt2_output) == 0 or 'generated_text' not in gpt2_output[0]: | |
| return None, "No se pudo generar el texto." | |
| def _progress(generated, to_generate): | |
| progress((generated, to_generate)) | |
| _progress(0, 1) | |
| try: | |
| tts = gTTS(generated_text, lang='es') | |
| temp_audio_path = "temp_audio.mp3" | |
| tts.save(temp_audio_path) | |
| audio_path = "audio.wav" | |
| audio = AudioSegment.from_mp3(temp_audio_path) | |
| audio.export(audio_path, format="wav") | |
| print("Archivo de audio generado:", audio_path) | |
| _progress(1, 2) | |
| except Exception as e: | |
| return None, f"No se pudo generar el audio: {str(e)}" | |
| command = f"python3 inference.py --checkpoint_path checkpoints/wav2lip_gan.pth --face oraculo.jpg --audio audio.wav --outfile video.mp4 --nosmooth" | |
| process = subprocess.run(command, shell=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE) | |
| if process.returncode != 0: | |
| error_message = process.stderr.decode("utf-8") | |
| return None, f"No se pudo generar el video: {error_message}" | |
| output_video_path = "video.mp4" | |
| os.remove(temp_audio_path) | |
| if os.path.isfile(output_video_path): | |
| _progress(2, 2) | |
| return output_video_path, None | |
| return None, "No se pudo generar el video" | |
| iface = gr.Interface( | |
| fn=generate_output, | |
| inputs=[ | |
| gr.inputs.Textbox(lines=1, label="Nombre", placeholder="Ingresa tu nombre"), | |
| gr.inputs.Textbox(lines=1, label="Fecha de Nacimiento", placeholder="DD/MM/AAAA") | |
| ], | |
| outputs=[ | |
| gr.outputs.Video(label="Respuesta de Andrea (un minuto aproximadamente)").style(width=256), | |
| #gr.outputs.Textbox(label="Mensaje de error", type="text") | |
| ], | |
| title="Oráculo de Inteligencia Artificial v2.1", | |
| description="Por favor, ingresa tu nombre y fecha de nacimiento." | |
| ) | |
| iface.launch() |