upload f5-tts source

app.py

@@ -1,8 +1,8 @@
-from flask import Flask, request, send_file, render_template
+from flask import Flask, request, send_file
 import subprocess
 import os
 import sys
-from io import BytesIO
+from huggingface_hub import hf_hub_download
 
 app = Flask(__name__)
 
@@ -12,10 +12,12 @@ app = Flask(__name__)
 def run_f5_tts(ref_audio_path, ref_text, gen_text, model="F5TTS_Base", speed=1.2, vocoder_name="vocos"):
     current_dir = os.path.dirname(os.path.abspath(__file__))
     infer_cli_path = os.path.join(current_dir, "src", "f5_tts", "infer", "infer_cli.py")
-    vocab_file = os.path.join(current_dir, "F5-TTS-MRSU", "vocab.txt")
-    ckpt_file = os.path.join(current_dir, "F5-TTS-MRSU", "model_last.pt")
     tests_dir = os.path.join(current_dir, "tests")
 
+    # Dùng huggingface_hub để tải file model và vocab từ repo 'nguyensu27/TTS'
+    vocab_file = hf_hub_download(repo_id="nguyensu27/TTS", filename="vocab.txt")
+    ckpt_file = hf_hub_download(repo_id="nguyensu27/TTS", filename="model_last.pt")
+
     os.environ['PYTHONIOENCODING'] = 'utf-8'
 
     command = [
@@ -43,7 +45,9 @@ def run_f5_tts(ref_audio_path, ref_text, gen_text, model="F5TTS_Base", speed=1.2
         if os.path.exists(tests_dir):
             wav_files = [f for f in os.listdir(tests_dir) if f.endswith('.wav')]
             if wav_files:
-                latest_wav = max(wav_files, key=lambda x: os.path.getmtime(os.path.join(tests_dir, x)))
+                latest_wav = max(
+                    wav_files, key=lambda x: os.path.getmtime(os.path.join(tests_dir, x))
+                )
                 output_file = os.path.join(tests_dir, latest_wav)
                 return True, output_file
 
@@ -53,14 +57,15 @@ def run_f5_tts(ref_audio_path, ref_text, gen_text, model="F5TTS_Base", speed=1.2
     except Exception as e:
         return False, str(e)
 
+
 # =========================
 # Routes
 # =========================
 @app.route('/')
 def home():
-    # Bạn có thể tạo index.html trong /templates nếu muốn giao diện
     return "F5-TTS API is running. Use POST /api/generate to generate audio."
 
+
 @app.route('/api/generate', methods=['POST'])
 def generate_speech():
     if 'ref_audio' not in request.files:
@@ -82,10 +87,10 @@ def generate_speech():
     else:
         return {"error": result}, 500
 
+
 # =========================
 # Main
 # =========================
 if __name__ == "__main__":
-    # Hugging Face Spaces tự set PORT
     port = int(os.environ.get("PORT", 7860))
     app.run(host="0.0.0.0", port=port, debug=False)

requirements.txt

@@ -39,4 +39,5 @@ tqdm
 transformers>=4.40
 uvicorn
 vocos
-x-transformers
+x-transformers
+huggingface_hub