Update app.py

app.py

@@ -24,6 +24,7 @@ import ffmpeg
 block = gr.Blocks()
 
 def calculate(image_in, audio_in):
+    print("in calculate")
     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)
@@ -40,38 +41,37 @@ def calculate(image_in, audio_in):
     return "/content/train/image_audio.mp4"
     
     
-def one_shot(image,input_text,gender): 
-   if gender == 'Female' or 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")
-          sound.export("/content/audio.wav", format="wav")
-     return calculate(image,"/content/audio.wav")
-   elif gender == 'Male' or gender == 'male':
-      print(gender)
-      models, cfg, task = load_model_ensemble_and_task_from_hf_hub(
-          "Voicemod/fastspeech2-en-male1",
+def one_shot(image,input_text,gender):
+    if gender == 'Female' or 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")
+            sound.export("/content/audio.wav", format="wav")
+        return calculate(image,"/content/audio.wav")
+    elif gender == 'Male' or gender == 'male':
+        print(gender)
+        models, cfg, task = load_model_ensemble_and_task_from_hf_hub(
+            "Voicemod/fastspeech2-en-male1",
           arg_overrides={"vocoder": "hifigan", "fp16": False}
-      )
-
-      model = models[0].cuda()
-      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"].cuda()
-      sample["net_input"]["src_lengths"] = sample["net_input"]["src_lengths"].cuda()
-      sample["speaker"] = sample["speaker"].cuda()
-
-      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)
-      calculate(image,'audio.wav')
+        )
+        model = models[0].cuda()
+        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"].cuda()
+        sample["net_input"]["src_lengths"] = sample["net_input"]["src_lengths"].cuda()
+        sample["speaker"] = sample["speaker"].cuda()
+        
+        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)
+        calculate(image,'audio.wav')