zeroGPU

app.py

@@ -2,6 +2,7 @@
 # Demo also available on HF Spaces: https://huggingface.co/spaces/mrfakename/MeloTTS
 import gradio as gr
 import os, torch, io
+import spaces
 os.system('python -m unidic download')
 # print("Make sure you've downloaded unidic (python -m unidic download) for this WebUI to work.")
 from melo.api import TTS
@@ -10,14 +11,19 @@ import tempfile
 import nltk
 nltk.download('averaged_perceptron_tagger_eng')
 device = 'cuda' if torch.cuda.is_available() else 'cpu'
-models = {
-    'EN': TTS(language='EN', device=device),
-    'ES': TTS(language='ES', device=device),
-    'FR': TTS(language='FR', device=device),
-    'ZH': TTS(language='ZH', device=device),
-    'JP': TTS(language='JP', device=device),
-    'KR': TTS(language='KR', device=device),
-}
+
+@spaces.GPU
+def load_models():
+    return {
+        'EN': TTS(language='EN', device=device),
+        'ES': TTS(language='ES', device=device),
+        'FR': TTS(language='FR', device=device),
+        'ZH': TTS(language='ZH', device=device),
+        'JP': TTS(language='JP', device=device),
+        'KR': TTS(language='KR', device=device),
+    }
+
+models = load_models()
 speaker_ids = models['EN'].hps.data.spk2id
 
 default_text_dict = {
@@ -29,6 +35,7 @@ default_text_dict = {
     'KR': '최근 텍스트 음성 변환 분야가 급속도로 발전하고 있습니다.',    
 }
     
+@spaces.GPU
 def synthesize(text, speaker, speed, language, progress=gr.Progress()):
     bio = io.BytesIO()
     models[language].tts_to_file(text, models[language].hps.data.spk2id[speaker], bio, speed=speed, pbar=progress.tqdm, format='wav')

requirements.txt

@@ -24,6 +24,7 @@ pypinyin==0.50.0
 cn2an==0.5.22
 jieba==0.42.1
 gradio
+spaces
 langid==1.1.6
 tqdm
 tensorboard==2.16.2