Update index.html

index.html

@@ -11,19 +11,31 @@ transformers_js_py
 </gradio-requirements>
 
 <gradio-file name="app.py" entrypoint>
-from transformers_js import import_transformers_js
+from transformers_js import pipeline
 import gradio as gr
+import numpy as np
+import scipy.io.wavfile as wavfile
 
 speaker_embeddings = 'https://huggingface.co/datasets/Xenova/transformers.js-docs/resolve/main/speaker_embeddings.bin';
 
-transformers = await import_transformers_js()
-pipeline = transformers.pipeline
-synthesizer = await pipeline('text-to-speech', 'Xenova/speecht5_tts', { "quantized": False })
+async def synthesize(text):
+    synthesizer = await pipeline(
+      'text-to-speech',
+      'Xenova/speecht5_tts',
+      { "quantized": False }
+    )  # Put the pipeline initializer inside the function to show the first view of the app faster
 
-async def classify(text):
-    return await synthesizer(text, { "speaker_embeddings": speaker_embeddings });
+    out = await synthesizer(text, { "speaker_embeddings": speaker_embeddings });
+    audio_data_memory_view = out["audio"]
+    sampling_rate = out["sampling_rate"]
 
-demo = gr.Interface(classify, "textbox", "audio")
+    audio_data = np.frombuffer(audio_data_memory_view, dtype=np.float32)
+
+    wavfile.write('output.wav', sampling_rate, audio_data)
+    return "output.wav"
+
+
+demo = gr.Interface(synthesize, "textbox", "file")
 demo.launch()
 </gradio-file>