Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -12,80 +12,55 @@ def load_eng_model():
|
|
| 12 |
model_path = hf_hub_download(repo_id=repo_id, filename="best_model.pth")
|
| 13 |
config_path = hf_hub_download(repo_id=repo_id, filename="config.json")
|
| 14 |
|
| 15 |
-
#
|
| 16 |
with open(config_path, 'r') as f:
|
| 17 |
-
|
| 18 |
|
| 19 |
-
# Force the character count in the dictionary
|
| 20 |
-
print("
|
| 21 |
-
|
| 22 |
|
| 23 |
-
# Save it back to
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
|
| 28 |
-
#
|
| 29 |
-
#
|
| 30 |
-
|
|
|
|
| 31 |
|
| 32 |
gc.collect()
|
| 33 |
return tts
|
| 34 |
|
| 35 |
# --- Global Initialization ---
|
| 36 |
-
print("Starting English TTS...")
|
| 37 |
try:
|
| 38 |
eng_tts = load_eng_model()
|
| 39 |
-
print("--- ENGLISH MODEL LOADED
|
| 40 |
except Exception as e:
|
| 41 |
print(f"CRITICAL ERROR: {e}")
|
| 42 |
eng_tts = None
|
| 43 |
|
| 44 |
def generate_voice(text):
|
| 45 |
-
print(f"\n>>> SYNTHESIS START: '{text}'")
|
| 46 |
if eng_tts is None:
|
| 47 |
-
print(">>> ERROR: eng_tts is None. Model did not load!")
|
| 48 |
return None
|
| 49 |
-
|
| 50 |
try:
|
| 51 |
-
# Use an absolute path to ensure Gradio finds it
|
| 52 |
output_path = os.path.join(os.getcwd(), "en_output.wav")
|
| 53 |
-
|
| 54 |
-
# Clean up old file if it exists
|
| 55 |
-
if os.path.exists(output_path):
|
| 56 |
-
os.remove(output_path)
|
| 57 |
-
print(">>> Cleaned old audio file.")
|
| 58 |
-
|
| 59 |
-
# Synthesis
|
| 60 |
-
print(">>> Running tts_to_file...")
|
| 61 |
eng_tts.tts_to_file(text=str(text), file_path=output_path)
|
| 62 |
-
|
| 63 |
-
# Check if the file actually exists and has data
|
| 64 |
-
if os.path.exists(output_path):
|
| 65 |
-
size = os.path.getsize(output_path)
|
| 66 |
-
if size > 1000: # Files smaller than 1KB are usually silent/empty
|
| 67 |
-
print(f">>> SUCCESS: Audio generated ({size} bytes)")
|
| 68 |
-
return output_path
|
| 69 |
-
else:
|
| 70 |
-
print(f">>> ERROR: File created but it is TOO SMALL ({size} bytes).")
|
| 71 |
-
else:
|
| 72 |
-
print(">>> ERROR: tts_to_file finished but NO FILE was found on disk.")
|
| 73 |
-
|
| 74 |
except Exception as e:
|
| 75 |
-
print(f"
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
return None
|
| 79 |
|
| 80 |
-
# Gradio Interface
|
| 81 |
demo = gr.Interface(
|
| 82 |
fn=generate_voice,
|
| 83 |
-
inputs=gr.Textbox(label="Input English Text"
|
| 84 |
-
outputs=gr.Audio(label="Synthesized Speech", type="filepath"
|
| 85 |
-
title="English VITS TTS"
|
| 86 |
-
allow_flagging="never"
|
| 87 |
)
|
| 88 |
|
| 89 |
-
|
| 90 |
if __name__ == "__main__":
|
| 91 |
demo.launch()
|
|
|
|
| 12 |
model_path = hf_hub_download(repo_id=repo_id, filename="best_model.pth")
|
| 13 |
config_path = hf_hub_download(repo_id=repo_id, filename="config.json")
|
| 14 |
|
| 15 |
+
# 1. Read the config file
|
| 16 |
with open(config_path, 'r') as f:
|
| 17 |
+
config_dict = json.load(f)
|
| 18 |
|
| 19 |
+
# 2. Force the character count to 137 in the dictionary
|
| 20 |
+
print("Forcing 137 characters in config...")
|
| 21 |
+
config_dict["model_args"]["num_chars"] = 137
|
| 22 |
|
| 23 |
+
# 3. Save it back to a NEW local file to ensure it's fresh
|
| 24 |
+
local_config_path = os.path.join(os.getcwd(), "fixed_config.json")
|
| 25 |
+
with open(local_config_path, 'w') as f:
|
| 26 |
+
json.dump(config_dict, f)
|
| 27 |
|
| 28 |
+
# 4. Initialize TTS and load model
|
| 29 |
+
# Use the local_config_path we just created
|
| 30 |
+
print("Initializing TTS engine with fixed config...")
|
| 31 |
+
tts = TTS(model_path=model_path, config_path=local_config_path, gpu=False)
|
| 32 |
|
| 33 |
gc.collect()
|
| 34 |
return tts
|
| 35 |
|
| 36 |
# --- Global Initialization ---
|
| 37 |
+
print("Starting English TTS Startup...")
|
| 38 |
try:
|
| 39 |
eng_tts = load_eng_model()
|
| 40 |
+
print("--- SUCCESS: ENGLISH MODEL LOADED ---")
|
| 41 |
except Exception as e:
|
| 42 |
print(f"CRITICAL ERROR: {e}")
|
| 43 |
eng_tts = None
|
| 44 |
|
| 45 |
def generate_voice(text):
|
|
|
|
| 46 |
if eng_tts is None:
|
|
|
|
| 47 |
return None
|
|
|
|
| 48 |
try:
|
|
|
|
| 49 |
output_path = os.path.join(os.getcwd(), "en_output.wav")
|
| 50 |
+
# Synthesize
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 51 |
eng_tts.tts_to_file(text=str(text), file_path=output_path)
|
| 52 |
+
return output_path
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
except Exception as e:
|
| 54 |
+
print(f"Generation Error: {e}")
|
| 55 |
+
return None
|
|
|
|
|
|
|
| 56 |
|
| 57 |
+
# Gradio Interface (Fixed: removed 'allow_flagging' which caused the crash)
|
| 58 |
demo = gr.Interface(
|
| 59 |
fn=generate_voice,
|
| 60 |
+
inputs=gr.Textbox(label="Input English Text"),
|
| 61 |
+
outputs=gr.Audio(label="Synthesized Speech", type="filepath"),
|
| 62 |
+
title="English VITS TTS"
|
|
|
|
| 63 |
)
|
| 64 |
|
|
|
|
| 65 |
if __name__ == "__main__":
|
| 66 |
demo.launch()
|