Update app.py

app.py

@@ -5,16 +5,15 @@ import tempfile
 from importlib.resources import files
 from silma_tts.api import SilmaTTS
 
+print("Loading SILMA TTS model...", flush=True)
+silma_tts = SilmaTTS()
+
 
 @spaces.GPU(duration=120)
 def do_inference(ref_audio, ref_text, gen_text):
     if not ref_audio:
         return None
 
-    print("Loading SILMA TTS model...", flush=True)
-    silma_tts = SilmaTTS()
-
-    
     _, output_wav_path = tempfile.mkstemp(suffix=".wav")
 
     wav, sr, _ = silma_tts.infer(
@@ -24,15 +23,24 @@ def do_inference(ref_audio, ref_text, gen_text):
         file_wave=output_wav_path,
     )
 
-    print("Done")
-    print(output_wav_path)
-    print(wav)
 
-    return wav
+    return output_wav_path
 
 
 with gr.Blocks() as demo:
-    gr.Markdown("# SILMA TTS Demo")
+
+        gr.HTML(
+        f"""<h1><img src='/gradio_api/file={CURR_BASE_DIR}/images/silma-logo.png'/>SILMA TTS Demo</h1>
+        <br>
+        <p style="font-size:16px">
+            SILMA TTS v1 is a high-performance, 150M-parameter bilingual (Arabic/English) TTS model developed by <a href="https://silma.ai">SILMA.AI</a>. You can download the model or learn more via the links below:
+            - <a href="https://huggingface.co/silma-ai/silma-tts" target="_blank">https://huggingface.co/silma-ai/silma-tts</a>
+            - <a href="https://github.com/SILMA-AI/silma-tts" target="_blank">https://github.com/SILMA-AI/silma-tts</a>      
+        </p>
+        """,
+        elem_id="tool-header"
+    )
+
     gr.Markdown("Add reference audio, reference text and generation text then click 'Generate Speech'")
     
     with gr.Row():