Update app.py

app.py

@@ -14,8 +14,8 @@ 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
+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()
 
@@ -126,8 +126,9 @@ 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):
@@ -141,7 +142,7 @@ def one_shot_talking(image_in,audio_in):
   # time.sleep(60)
   image_in_one_shot='/content/results/restored_imgs/image_pre.png'
   #One Shot Talking Face algorithm
-  return calculate(image_in_one_shot,audio_in)
+  calculate(image_in_one_shot,audio_in)
 
   #Video Quality Improvement
 
@@ -169,7 +170,31 @@ def one_shot(image_in,input_text,gender):
               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)
+          soundfile.write("/content/audio_before.wav", wav.cpu().clone().numpy(), rate)
+          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():
@@ -179,7 +204,7 @@ def run():
       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=False,label="Text")
+          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):