Update app.py

app.py

@@ -23,12 +23,33 @@ import ffmpeg
 
 block = gr.Blocks()
 
-def cal(gender,input_text):
-    return gender+input_text
+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)
+    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"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 one_shot(image,input_text,gender): 
    if gender == 'Female' or gender == 'female':
-     return cal(gender,input_text)
+       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")
+           sound.export("/content/audio.wav", format="wav")
+           audio_in="/content/audio.wav"
+     return calculate(image_in,audio_in)
      
        
 
@@ -46,7 +67,7 @@ def run():
           # audio_in = gr.Audio(show_label=False, type='filepath')
           input_text=gr.Textbox(lines=3, value="Hello How are you?", label="Input Text")
           gender = gr.Radio(["Female","Male"],value="Female",label="Gender")
-          video_out = gr.Textbox(label="output")
+          video_out = gr.Video(label="output")
           # video_out = gr.Video(show_label=False)
         with gr.Row().style(equal_height=True):
           btn = gr.Button("Generate")