Update app.py

app.py

@@ -6,6 +6,8 @@ from transformers import BlipProcessor, BlipForConditionalGeneration
 from fairseq.checkpoint_utils import load_model_ensemble_and_task_from_hf_hub
 from fairseq.models.text_to_speech.hub_interface import TTSHubInterface
 import IPython.display as ipd
+import requests
+
 
 
 processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
@@ -34,20 +36,14 @@ with gr.Blocks(theme=gr.themes.Ocean(primary_hue="pink", neutral_hue="indigo", f
       return processor.decode(out[0], skip_special_tokens=True)
 
     def leer(texto):
-        models, cfg, task = load_model_ensemble_and_task_from_hf_hub(
-        "facebook/fastspeech2-en-ljspeech",
-        arg_overrides={"vocoder": "hifigan", "fp16": False}
-        )
-        modelA = models[0]
-        TTSHubInterface.update_cfg_with_data_cfg(cfg, task.data_cfg)
-        generator = task.build_generator(modelA, cfg)
-        
-        text = texto
+        response = requests.post("https://charly-text-to-speech.hf.space/run/predict", json={
+          "data": [
+            texto,
+        ]}).json()
         
-        sample = TTSHubInterface.get_model_input(task, text)
-        wav, rate = TTSHubInterface.get_prediction(task, modelA, generator, sample)
+        data = response["data"]
         
-        return ipd.Audio(wav, rate=rate)
+        return data
 
 
     button.click(describir, [textbox], output)