app.py

import gradio as gr
import os, subprocess, torchaudio
import torch
from PIL import Image
import gradio as gr
import soundfile
from gtts import gTTS
import tempfile
from pydub import AudioSegment
from pydub.generators import Sine
import dlib
import cv2
import imageio
import os
import ffmpeg

from fairseq.checkpoint_utils import load_model_ensemble_and_task_from_hf_hub
from fairseq.models.text_to_speech.hub_interface import TTSHubInterface

block = gr.Blocks()

def compute_aspect_preserved_bbox(bbox, increase_area, h, w):
    left, top, right, bot = bbox
    width = right - left
    height = bot - top

    width_increase = max(increase_area, ((1 + 2 * increase_area) * height - width) / (2 * width))
    height_increase = max(increase_area, ((1 + 2 * increase_area) * width - height) / (2 * height))

    left_t = int(left - width_increase * width)
    top_t = int(top - height_increase * height)
    right_t = int(right + width_increase * width)
    bot_t = int(bot + height_increase * height)

    left_oob = -min(0, left_t)
    right_oob = right - min(right_t, w)
    top_oob = -min(0, top_t)
    bot_oob = bot - min(bot_t, h)

    if max(left_oob, right_oob, top_oob, bot_oob) > 0:
        max_w = max(left_oob, right_oob)
        max_h = max(top_oob, bot_oob)
        if max_w > max_h:
            return left_t + max_w, top_t + max_w, right_t - max_w, bot_t - max_w
        else:
            return left_t + max_h, top_t + max_h, right_t - max_h, bot_t - max_h

    else:
        return (left_t, top_t, right_t, bot_t)

def crop_src_image(src_img, detector=None):
    if  detector is None:
        detector = dlib.get_frontal_face_detector()
    save_img='/content/image_pre.png'
    img = cv2.imread(src_img)
    faces = detector(img, 0)
    h, width, _ = img.shape
    if len(faces) > 0:
        bbox = [faces[0].left(), faces[0].top(),faces[0].right(), faces[0].bottom()]
        l = bbox[3]-bbox[1]
        bbox[1]= bbox[1]-l*0.1
        bbox[3]= bbox[3]-l*0.1
        bbox[1] = max(0,bbox[1])
        bbox[3] = min(h,bbox[3])
        bbox = compute_aspect_preserved_bbox(tuple(bbox), 0.5, img.shape[0], img.shape[1])
        img = img[bbox[1] :bbox[3] , bbox[0]:bbox[2]]
        img = cv2.resize(img, (256, 256))
        cv2.imwrite(save_img,img)
    else:
        img = cv2.resize(img,(256,256))
        cv2.imwrite(save_img, img)
    return save_img

def pad_image(image):
    w, h = image.size
    if w == h:
        return image
    elif w > h:
        new_image = Image.new(image.mode, (w, w), (0, 0, 0))
        new_image.paste(image, (0, (w - h) // 2))
        return new_image
    else:
        new_image = Image.new(image.mode, (h, h), (0, 0, 0))
        new_image.paste(image, ((h - w) // 2, 0))
        return new_image

def calculate(image_in, audio_in):
    waveform, sample_rate = torchaudio.load(audio_in)
    waveform = torch.mean(waveform, dim=0, keepdim=True)
    torchaudio.save("/content/audio.wav", waveform, sample_rate, encoding="PCM_S", bits_per_sample=16)
    image = Image.open(image_in)
    image = pad_image(image)
    # os.system(f"rm -rf /content/image.png")
    image.save("image.png")

    pocketsphinx_run = subprocess.run(['pocketsphinx', '-phone_align', 'yes', 'single', '/content/audio.wav'], check=True, capture_output=True)
    jq_run = subprocess.run(['jq', '[.w[]|{word: (.t | ascii_upcase | sub("<S>"; "sil") | sub("<SIL>"; "sil") | sub("\\\(2\\\)"; "") | sub("\\\(3\\\)"; "") | sub("\\\(4\\\)"; "") | sub("\\\[SPEECH\\\]"; "SIL") | sub("\\\[NOISE\\\]"; "SIL")), phones: [.w[]|{ph: .t | sub("\\\+SPN\\\+"; "SIL") | sub("\\\+NSN\\\+"; "SIL"), bg: (.b*100)|floor, ed: (.b*100+.d*100)|floor}]}]'], input=pocketsphinx_run.stdout, capture_output=True)
    with open("test.json", "w") as f:
        f.write(jq_run.stdout.decode('utf-8').strip())
    # device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
    os.system(f"rm -rf /content/image_audio.mp4")
    os.system(f"cd /content/one-shot-talking-face && python3 -B test_script.py --img_path /content/image.png --audio_path /content/audio.wav --phoneme_path /content/test.json --save_dir /content/train")
    return "/content/train/image_audio.mp4"

def merge_frames():


  path = '/content/video_results/restored_imgs'
  image_folder = os.fsencode(path)
  print(image_folder)
  filenames = []

  for file in os.listdir(image_folder):
      filename = os.fsdecode(file)
      if filename.endswith( ('.jpg', '.png', '.gif') ):
          filenames.append(filename)

  filenames.sort() # this iteration technique has no built in order, so sort the frames
  print(filenames)
  images = list(map(lambda filename: imageio.imread("/content/video_results/restored_imgs/"+filename), filenames))
  os.system(f"rm -rf /content/video_output.mp4")
  imageio.mimsave('/content/video_output.mp4', images, fps=25.0) # modify the frame duration as needed
  return "/content/video_output.mp4"

def audio_video():

  input_video = ffmpeg.input('/content/video_output.mp4')

  input_audio = ffmpeg.input('/content/audio.wav')
  os.system(f"rm -rf /content/final_output.mp4")
  ffmpeg.concat(input_video, input_audio, v=1, a=1).output('/content/final_output.mp4').run()
  
  return "/content/final_output.mp4"

def one_shot_talking(image_in,audio_in):


  # Pre-processing of image
  crop_img=crop_src_image(image_in)

  os.system(f"rm -rf /content/results/restored_imgs/image_pre.png")
  #Improve quality of input image
  os.system(f"python /content/GFPGAN/inference_gfpgan.py --upscale 2 -i /content/image_pre.png -o /content/results  --bg_upsampler realesrgan")
  # time.sleep(60)
  
  image_in_one_shot='/content/results/restored_imgs/image_pre.png'
  
  #One Shot Talking Face algorithm
  calculate(image_in_one_shot,audio_in)

  #Video Quality Improvement

  #1. Extract the frames from the video file using PyVideoFramesExtractor
  os.system(f"python /content/PyVideoFramesExtractor/extract.py --video=/content/train/image_audio.mp4")

  #2. Improve image quality using GFPGAN on each frames
  os.system(f"rm -rf /content/extracted_frames/image_audio_frames")
  os.system(f"rm -rf /content/video_results/")
  os.system(f"python /content/GFPGAN/inference_gfpgan.py --upscale 2 -i /content/extracted_frames/image_audio_frames -o /content/video_results  --bg_upsampler realesrgan")

  #3. Merge all the frames to a one video using imageio
  merge_frames()
  return audio_video()


def one_shot(image_in,input_text,gender):
    if gender == "Female":
         tts = gTTS(input_text)
         with tempfile.NamedTemporaryFile(suffix='.mp3', delete=False) as f:
              tts.write_to_fp(f)
              f.seek(0)
              sound = AudioSegment.from_file(f.name, format="mp3")
              os.system(f"rm -rf /content/audio.wav")
              sound.export("/content/audio.wav", format="wav")
              audio_in="/content/audio.wav"
         return one_shot_talking(image_in,audio_in)
    elif gender == 'Male':
       
          models, cfg, task = load_model_ensemble_and_task_from_hf_hub(
              "Voicemod/fastspeech2-en-male1",
              arg_overrides={"vocoder": "hifigan", "fp16": False}
          )

          model = models[0]
          TTSHubInterface.update_cfg_with_data_cfg(cfg, task.data_cfg)
          generator = task.build_generator([model], cfg)
          # next(model.parameters()).device

          sample = TTSHubInterface.get_model_input(task, input_text)
          sample["net_input"]["src_tokens"] = sample["net_input"]["src_tokens"]
          sample["net_input"]["src_lengths"] = sample["net_input"]["src_lengths"]
          sample["speaker"] = sample["speaker"]

          wav, rate = TTSHubInterface.get_prediction(task, model, generator, sample)
          # soundfile.write("/content/audio_before.wav", wav, rate)
          os.system(f"rm -rf /content/audio_before.wav")
          soundfile.write("/content/audio_before.wav", wav.cpu().clone().numpy(), rate)
          os.system(f"rm -rf /content/audio.wav")
          cmd='ffmpeg -i /content/audio_before.wav -filter:a "atempo=0.7" -vn /content/audio.wav'
          os.system(cmd)
          audio_in="/content/audio.wav"
          
          return one_shot_talking(image_in,audio_in)
           
                        
def run():
  with gr.Blocks(css=".gradio-container {background-color: lightgray} #radio_div {background-color: #FFD8B4; font-size: 40px;}") as demo:
    gr.Markdown("<h1 style='text-align: center;'>"+ "One Shot Talking Face from Text" + "</h1><br/><br/>")
    with gr.Group():
      with gr.Box():
        with gr.Row().style(equal_height=True):
          image_in = gr.Image(show_label=True, type="filepath",label="Input Image")
          input_text = gr.Textbox(show_label=True,label="Input Text")
          gender = gr.Radio(["Female","Male"],value="Female",label="Gender")
          video_out = gr.Video(show_label=True,label="Output")
        with gr.Row().style(equal_height=True):
          btn = gr.Button("Generate")   
    gr.Markdown(
            """
            <p style='text-align: center;'>Feel free to give us your thoughts on this demo and please contact us at 
                    <a href="mailto:letstalk@pragnakalp.com" target="_blank">letstalk@pragnakalp.com</a> 
                    <p style='text-align: center;'>Developed by: <a href="https://www.pragnakalp.com" target="_blank">Pragnakalp Techlabs</a></p>

            """)

    btn.click(one_shot, inputs=[image_in,input_text,gender], outputs=[video_out])
    demo.queue()
    demo.launch(server_name="0.0.0.0", server_port=7860)


if __name__ == "__main__":
    run()