Update app.py

app.py

@@ -25,20 +25,7 @@ tts_model_collections = {
 }
 
 
-def gpu_decorator(func):
-    return spaces.GPU(func)
-
-
-def load_text_from_file(file):
-    if file:
-        with open(file, "r", encoding="utf-8") as f:
-            text = f.read().strip()
-    else:
-        text = ""
-    return gr.update(value=text)
-
-
-@gpu_decorator
+@spaces.GPU
 def infer(
     ref_audio_orig,
     ref_text,
@@ -51,9 +38,9 @@ def infer(
         gr.Warning("Please ensure [Reference Audio] [Reference Text] [Text to Generate] are all provided.")
         return gr.update(), gr.update(), ref_text
 
-    if seed < 0 or seed > 1024:
-        gr.Warning("Please set a seed in range 0 ~ 1024.")
-        seed = np.random.randint(0, 1024)
+    if seed < 0 or seed > 2**31 - 1:
+        gr.Warning("Please set a seed in range 0 ~ 2**31 - 1.")
+        seed = np.random.randint(0, 2**31 - 1)
     torch.manual_seed(seed)
     used_seed = seed
 
@@ -74,22 +61,22 @@ def infer(
 
 with gr.Blocks() as app_basic_tts:
     gr.Markdown("# Batched TTS")
-    ref_wav_input = gr.Audio(label="Reference Audio", type="filepath")
-    ref_txt_input = gr.Textbox(label="Reference Text")
-    gen_txt_input = gr.Textbox(label="Text to Generate")
-    generate_btn = gr.Button("Synthesize", variant="primary")
     with gr.Row():
-        randomize_seed = gr.Checkbox(
-            label="Randomize Seed",
-            info="Check to use a random seed for each generation. Uncheck to use the seed specified.",
-            value=True,
-            scale=3,
-        )
-        seed_input = gr.Number(show_label=False, value=0, precision=0, scale=1)
-
-    audio_output = gr.Audio(label="Synthesized Audio")
+        with gr.Column():
+            ref_wav_input = gr.Audio(label="Reference Audio", type="filepath")
+            ref_txt_input = gr.Textbox(label="Reference Text")
+            gen_txt_input = gr.Textbox(label="Text to Generate")
+            generate_btn = gr.Button("Synthesize", variant="primary")
+            with gr.Row():
+                randomize_seed = gr.Checkbox(
+                    label="Randomize Seed",
+                    info="Check to use a random seed for each generation. Uncheck to use the seed specified.",
+                    value=True,
+                    scale=3,
+                )
+                seed_input = gr.Number(show_label=False, value=0, precision=0, scale=1)
+        audio_output = gr.Audio(label="Synthesized Audio")
 
-    @gpu_decorator
     def basic_tts(
         ref_wav_input,
         ref_txt_input,
@@ -98,7 +85,7 @@ with gr.Blocks() as app_basic_tts:
         seed_input,
     ):
         if randomize_seed:
-            seed_input = np.random.randint(0, 1024)
+            seed_input = np.random.randint(0, 2**31 - 1)
 
         audio_out, ref_text_out, used_seed = infer(
             ref_wav_input,
@@ -110,7 +97,7 @@ with gr.Blocks() as app_basic_tts:
         return audio_out, ref_text_out, used_seed
 
     ref_wav_input.clear(
-        lambda: [None, None],
+        lambda: [None],
         None,
         [ref_txt_input],
     )
@@ -131,7 +118,9 @@ with gr.Blocks() as app_basic_tts:
 with gr.Blocks() as demo:
     gr.Markdown(
         """
-        # F5-TTS
+        # 🗣️ F5-TTS Online Demo for Dev Test
+        
+        Upload/record a reference voice, give reference and generation text, and enjoy playing!
         """
     )