Update app.py

app.py

@@ -2,7 +2,6 @@ import spaces
 from kokoro import KModel, KPipeline
 import gradio as gr
 import os
-import random
 import torch
 
 IS_DUPLICATE = not os.getenv('SPACE_ID', '').startswith('hexgrad/')
@@ -44,33 +43,6 @@ def generate_first(text, voice='af_heart', speed=1, use_gpu=CUDA_AVAILABLE):
 def predict(text, voice='af_heart', speed=1):
     return generate_first(text, voice, speed, use_gpu=False)[0]
 
-def tokenize_first(text, voice='af_heart'):
-    pipeline = pipelines[voice[0]]
-    for _, ps, _ in pipeline(text, voice):
-        return ps
-    return ''
-
-def generate_all(text, voice='af_heart', speed=1, use_gpu=CUDA_AVAILABLE):
-    text = text if CHAR_LIMIT is None else text.strip()[:CHAR_LIMIT]
-    pipeline = pipelines[voice[0]]
-    pack = pipeline.load_voice(voice)
-    use_gpu = use_gpu and CUDA_AVAILABLE
-    for _, ps, _ in pipeline(text, voice, speed):
-        ref_s = pack[len(ps)-1]
-        try:
-            if use_gpu:
-                audio = forward_gpu(ps, ref_s, speed)
-            else:
-                audio = models[False](ps, ref_s, speed)
-        except gr.exceptions.Error as e:
-            if use_gpu:
-                gr.Warning(str(e))
-                gr.Info('Switching to CPU')
-                audio = models[False](ps, ref_s, speed)
-            else:
-                raise gr.Error(e)
-        yield 24000, audio.numpy()
-
 random_texts = {}
 for lang in ['en']:
     with open(f'{lang}.txt', 'r') as r:
@@ -92,7 +64,7 @@ CHOICES = {
 '🇺🇸 🚺 Alloy': 'af_alloy',
 '🇺🇸 🚺 Jessica': 'af_jessica',
 '🇺🇸 🚺 River': 'af_river',
-    
+
 '🇺🇸 🚹 Michael': 'am_michael',
 '🇺🇸 🚹 Fenrir': 'am_fenrir',
 '🇺🇸 🚹 Puck': 'am_puck',
@@ -102,46 +74,46 @@ CHOICES = {
 '🇺🇸 🚹 Onyx': 'am_onyx',
 '🇺🇸 🚹 Santa': 'am_santa',
 '🇺🇸 🚹 Adam': 'am_adam',
-    
+
 '🇬🇧 🚺 Emma': 'bf_emma',
 '🇬🇧 🚺 Isabella': 'bf_isabella',
 '🇬🇧 🚺 Alice': 'bf_alice',
 '🇬🇧 🚺 Lily': 'bf_lily',
-    
+
 '🇬🇧 🚹 George': 'bm_george',
 '🇬🇧 🚹 Fable': 'bm_fable',
 '🇬🇧 🚹 Lewis': 'bm_lewis',
 '🇬🇧 🚹 Daniel': 'bm_daniel',
-    
+
 '🇪🇸 🚺 Dora': 'ef_dora',
-    
+
 '🇪🇸 🚹 Alex': 'em_alex',
 '🇪🇸 🚹 Santa': 'em_santa',
-    
+
 '🇫🇷 🚺 Siwis': 'ff_siwis',
-    
+
 '🇮🇳 🚹 Alpha': 'hf_alpha',
 '🇮🇳 🚹 Beta': 'hf_beta',
-    
+
 '🇮🇳 🚹 Omega': 'hm_omega',
 '🇮🇳 🚹 Psi': 'hm_psi',
-    
+
 '🇮🇹 🚺 Sara': 'if_sara',
-    
+
 '🇮🇹 🚺 Nicola': 'im_nicola',
-    
+
 '🇯🇵 🚹 Alpha': 'jf_alpha',
 '🇯🇵 🚹 Gongitsune': 'jf_gongitsune',
 '🇯🇵 🚹 Nezumi': 'jf_nezumi',
 '🇯🇵 🚹 Tebukuro': 'jf_tebukuro',
-    
+
 '🇯🇵 🚹 Kumo': 'jm_kumo',
-    
+
 '🇧🇷 🚺 Dora': 'pf_dora',
-    
+
 '🇧🇷 🚹 Alex': 'pm_alex',
 '🇧🇷 🚹 Santa': 'pm_santa',
-    
+
 '🇨🇳 🚺 Xiaobei': 'zf_xiaobei',
 '🇨🇳 🚺 Xiaoni': 'zf_xiaoni',
 '🇨🇳 🚺 Xiaoxiao': 'zf_xiaoxiao',
@@ -155,69 +127,28 @@ CHOICES = {
 for v in CHOICES.values():
     pipelines[v[0]].load_voice(v)
 
-TOKEN_NOTE = '''
-💡 Customize pronunciation with Markdown link syntax and /slashes/ like [Kokoro](/kˈOkəɹO/)
-💬 To adjust intonation, try punctuation ;:,.!?—…"()“” or stress ˈ and ˌ
-⬇️ Lower stress [1 level](-1) or [2 levels](-2)
-⬆️ Raise stress 1 level [or](+2) 2 levels (only works on less stressed, usually short words)
-'''
-
-with gr.Blocks() as generate_tab:
-    out_audio = gr.Audio(label='Output Audio', interactive=False, streaming=False, autoplay=True)
-    generate_btn = gr.Button('Generate', variant='primary')
-    with gr.Accordion('Output Tokens', open=True):
-        out_ps = gr.Textbox(interactive=False, show_label=False, info='Tokens used to generate the audio, up to 510 context length.')
-        tokenize_btn = gr.Button('Tokenize', variant='secondary')
-        gr.Markdown(TOKEN_NOTE)
-        predict_btn = gr.Button('Predict', variant='secondary', visible=False)
-
-STREAM_NOTE = ['⚠️ There is an unknown Gradio bug that might yield no audio the first time you click Stream.']
-if CHAR_LIMIT is not None:
-    STREAM_NOTE.append(f'✂️ Each stream is capped at {CHAR_LIMIT} characters.')
-    STREAM_NOTE.append('🚀 Want more characters? You can [use Kokoro directly](https://huggingface.co/hexgrad/Kokoro-82M#usage) or duplicate this space:')
-STREAM_NOTE = '\n\n'.join(STREAM_NOTE)
-
-with gr.Blocks() as stream_tab:
-    out_stream = gr.Audio(label='Output Audio Stream', interactive=False, streaming=True, autoplay=True)
-    with gr.Row():
-        stream_btn = gr.Button('Stream', variant='primary')
-        stop_btn = gr.Button('Stop', variant='stop')
-    with gr.Accordion('Note', open=True):
-        gr.Markdown(STREAM_NOTE)
-        gr.DuplicateButton()
-
 BANNER_TEXT = '''
 [***Kokoro*** **is an open-weight TTS model with 82 million parameters.**](https://huggingface.co/hexgrad/Kokoro-82M)
 As of January 31st, 2025, Kokoro was the most-liked [**TTS model**](https://huggingface.co/models?pipeline_tag=text-to-speech&sort=likes) and the most-liked [**TTS space**](https://huggingface.co/spaces?sort=likes&search=tts) on Hugging Face.
 This demo only showcases English, but you can directly use the model to access other languages.
 '''
-API_OPEN = os.getenv('SPACE_ID') != 'hexgrad/Kokoro-TTS'
-API_NAME = None if API_OPEN else False
+
 with gr.Blocks() as app:
     with gr.Row():
         gr.Markdown(BANNER_TEXT, container=True)
     with gr.Row():
         with gr.Column():
-            text = gr.Textbox(label='Input Text', info=f"Up to ~500 characters per Generate, or {'∞' if CHAR_LIMIT is None else CHAR_LIMIT} characters per Stream")
+            text = gr.Textbox(label='Input Text', info=f"Up to ~500 characters per Generate")
             with gr.Row():
                 voice = gr.Dropdown(list(CHOICES.items()), value='af_heart', label='Voice', info='Quality and availability vary by language')
-                use_gpu = gr.Dropdown(
-                    [('ZeroGPU 🚀', True), ('CPU 🐌', False)],
-                    value=CUDA_AVAILABLE,
-                    label='Hardware',
-                    info='GPU is usually faster, but has a usage quota',
-                    interactive=CUDA_AVAILABLE
-                )
             speed = gr.Slider(minimum=0.5, maximum=2, value=1, step=0.1, label='Speed')
             random_btn = gr.Button('Random Text', variant='secondary')
         with gr.Column():
-            gr.TabbedInterface([generate_tab, stream_tab], ['Generate', 'Stream'])
-    random_btn.click(fn=get_random_text, inputs=[voice], outputs=[text], api_name=API_NAME)
-    generate_btn.click(fn=generate_first, inputs=[text, voice, speed, use_gpu], outputs=[out_audio, out_ps], api_name=API_NAME)
-    tokenize_btn.click(fn=tokenize_first, inputs=[text, voice], outputs=[out_ps], api_name=API_NAME)
-    stream_event = stream_btn.click(fn=generate_all, inputs=[text, voice, speed, use_gpu], outputs=[out_stream], api_name=API_NAME)
-    stop_btn.click(fn=None, cancels=stream_event)
-    predict_btn.click(fn=predict, inputs=[text, voice, speed], outputs=[out_audio], api_name=API_NAME)
+            out_audio = gr.Audio(label='Output Audio', interactive=False, streaming=False, autoplay=True)
+            generate_btn = gr.Button('Generate', variant='primary')
+            
+    random_btn.click(fn=get_random_text, inputs=[voice], outputs=[text])
+    generate_btn.click(fn=generate_first, inputs=[text, voice, speed], outputs=[out_audio], api_name=None)
 
 if __name__ == '__main__':
-    app.queue(api_open=API_OPEN).launch(show_api=API_OPEN, ssr_mode=True)
+    app.queue().launch()