Spaces:
Running
on
Zero
Running
on
Zero
Fix error: generate_speech takes 6 arguments
Browse files
app.py
CHANGED
|
@@ -60,7 +60,7 @@ print("All models loaded!")
|
|
| 60 |
|
| 61 |
|
| 62 |
@spaces.GPU
|
| 63 |
-
def generate_speech_gpu(text, model_choice):
|
| 64 |
"""
|
| 65 |
Generate speech from text using the selected model on GPU
|
| 66 |
"""
|
|
@@ -102,8 +102,8 @@ with gr.Blocks(title="π» KaniTTS - Text to Speech", theme=gr.themes.Default())
|
|
| 102 |
model_dropdown = gr.Dropdown(
|
| 103 |
choices=list(models_configs.keys()),
|
| 104 |
value=list(models_configs.keys())[0],
|
| 105 |
-
label="
|
| 106 |
-
info="Base
|
| 107 |
)
|
| 108 |
|
| 109 |
text_input = gr.Textbox(
|
|
@@ -112,6 +112,28 @@ with gr.Blocks(title="π» KaniTTS - Text to Speech", theme=gr.themes.Default())
|
|
| 112 |
lines=3,
|
| 113 |
max_lines=10
|
| 114 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 115 |
|
| 116 |
generate_btn = gr.Button("π΅ Generate Speech", variant="primary", size="lg")
|
| 117 |
|
|
@@ -128,28 +150,6 @@ with gr.Blocks(title="π» KaniTTS - Text to Speech", theme=gr.themes.Default())
|
|
| 128 |
value="Ready to generate speech",
|
| 129 |
lines=3
|
| 130 |
)
|
| 131 |
-
|
| 132 |
-
with gr.Accordion("Settings", open=False):
|
| 133 |
-
temperature = gr.Slider(
|
| 134 |
-
minimum=0.1, maximum=1.5, value=0.6, step=0.05,
|
| 135 |
-
label="Temperature",
|
| 136 |
-
info="Higher values (0.7-1.0) create more expressive but less stable speech"
|
| 137 |
-
)
|
| 138 |
-
top_p = gr.Slider(
|
| 139 |
-
minimum=0.1, maximum=1.0, value=0.95, step=0.05,
|
| 140 |
-
label="Top P",
|
| 141 |
-
info="Nucleus sampling threshold"
|
| 142 |
-
)
|
| 143 |
-
repetition_penalty = gr.Slider(
|
| 144 |
-
minimum=1.0, maximum=2.0, value=1.1, step=0.05,
|
| 145 |
-
label="Repetition Penalty",
|
| 146 |
-
info="Higher values discourage repetitive patterns"
|
| 147 |
-
)
|
| 148 |
-
max_new_tokens = gr.Slider(
|
| 149 |
-
minimum=100, maximum=2000, value=1200, step=100,
|
| 150 |
-
label="Max Length",
|
| 151 |
-
info="Maximum length of generated audio (in tokens)"
|
| 152 |
-
)
|
| 153 |
|
| 154 |
# GPU generation event
|
| 155 |
generate_btn.click(
|
|
@@ -158,7 +158,6 @@ with gr.Blocks(title="π» KaniTTS - Text to Speech", theme=gr.themes.Default())
|
|
| 158 |
outputs=[audio_output, time_report_output]
|
| 159 |
)
|
| 160 |
|
| 161 |
-
gr.Markdown("## Examples")
|
| 162 |
|
| 163 |
def play_demo(text):
|
| 164 |
return (22050, demo_examples[text]), 'DEMO'
|
|
|
|
| 60 |
|
| 61 |
|
| 62 |
@spaces.GPU
|
| 63 |
+
def generate_speech_gpu(text, model_choice, temperature, top_p, repetition_penalty, max_new_tokens):
|
| 64 |
"""
|
| 65 |
Generate speech from text using the selected model on GPU
|
| 66 |
"""
|
|
|
|
| 102 |
model_dropdown = gr.Dropdown(
|
| 103 |
choices=list(models_configs.keys()),
|
| 104 |
value=list(models_configs.keys())[0],
|
| 105 |
+
label="Selected Model",
|
| 106 |
+
info="Base generates random voices"
|
| 107 |
)
|
| 108 |
|
| 109 |
text_input = gr.Textbox(
|
|
|
|
| 112 |
lines=3,
|
| 113 |
max_lines=10
|
| 114 |
)
|
| 115 |
+
|
| 116 |
+
with gr.Accordion("Settings", open=False):
|
| 117 |
+
temperature = gr.Slider(
|
| 118 |
+
minimum=0.1, maximum=1.5, value=0.6, step=0.05,
|
| 119 |
+
label="Temperature",
|
| 120 |
+
info="Higher values (0.7-1.0) create more expressive but less stable speech"
|
| 121 |
+
)
|
| 122 |
+
top_p = gr.Slider(
|
| 123 |
+
minimum=0.1, maximum=1.0, value=0.95, step=0.05,
|
| 124 |
+
label="Top P",
|
| 125 |
+
info="Nucleus sampling threshold"
|
| 126 |
+
)
|
| 127 |
+
repetition_penalty = gr.Slider(
|
| 128 |
+
minimum=1.0, maximum=2.0, value=1.1, step=0.05,
|
| 129 |
+
label="Repetition Penalty",
|
| 130 |
+
info="Higher values discourage repetitive patterns"
|
| 131 |
+
)
|
| 132 |
+
max_new_tokens = gr.Slider(
|
| 133 |
+
minimum=100, maximum=2000, value=1200, step=100,
|
| 134 |
+
label="Max Length",
|
| 135 |
+
info="Maximum length of generated audio (in tokens)"
|
| 136 |
+
)
|
| 137 |
|
| 138 |
generate_btn = gr.Button("π΅ Generate Speech", variant="primary", size="lg")
|
| 139 |
|
|
|
|
| 150 |
value="Ready to generate speech",
|
| 151 |
lines=3
|
| 152 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 153 |
|
| 154 |
# GPU generation event
|
| 155 |
generate_btn.click(
|
|
|
|
| 158 |
outputs=[audio_output, time_report_output]
|
| 159 |
)
|
| 160 |
|
|
|
|
| 161 |
|
| 162 |
def play_demo(text):
|
| 163 |
return (22050, demo_examples[text]), 'DEMO'
|