Fix speaker2refs parsing (dict) and ref path

app.py

@@ -44,12 +44,11 @@ SPEAKER_CHOICES = sorted(SPEAKER2REFS.keys())
 if not SPEAKER_CHOICES:
     raise RuntimeError("speaker2refs.json is empty (no speakers found).")
 
-# def _abs_audio(p: str) -> str:
-#     return p if os.path.isabs(p) else os.path.join(DATA_ROOT, p)
 
 def _abs_audio(p: str) -> str:
     # p trong json đang là "demo_data/refs/id_k.wav" => join theo repo root
     return p if os.path.isabs(p) else os.path.join(repo_dir, p)
+
 # =========================
 # LOAD MODEL
 # =========================
@@ -73,60 +72,6 @@ def _cache_set(key, val):
         STYLE_CACHE.pop(next(iter(STYLE_CACHE)))
     STYLE_CACHE[key] = val
 
-# @torch.inference_mode()
-# def synth_one_speaker(speaker_name: str, text_prompt: str,
-#                       denoise: float, avg_style: bool, stabilize: bool):
-#     try:
-#         if not speaker_name:
-#             return None, "Bạn chưa chọn speaker."
-
-#         refs = SPEAKER2REFS.get(speaker_name, [])
-#         if not refs:
-#             return None, f"Speaker '{speaker_name}' không có ref trong speaker2refs.json."
-
-#         ref_path = _abs_audio(refs[0])
-#         if not os.path.isfile(ref_path):
-#             return None, f"Ref audio not found: {ref_path}"
-
-#         if not text_prompt or not text_prompt.strip():
-#             return None, "Bạn chưa nhập text."
-
-#         speakers = {
-#             "id_1": {"path": ref_path, "lang": "vi", "speed": 1.0}
-#         }
-
-#         cache_key = (speaker_name, float(denoise), bool(avg_style))
-#         styles = _cache_get(cache_key)
-#         if styles is None:
-#             styles = model.get_styles(speakers, denoise, avg_style)
-#             _cache_set(cache_key, styles)
-
-#         # Neu user khong them tag [id_1] thi tu them
-#         text_prompt = text_prompt.strip()
-#         if "[id_" not in text_prompt:
-#             text_prompt = "[id_1] " + text_prompt
-
-#         r = model.generate(text_prompt, styles, stabilize, 18, "[id_1]")
-
-#         r = np.asarray(r, dtype=np.float32)
-#         m = float(np.max(np.abs(r))) if r.size else 0.0
-#         if m > 1e-9:
-#             r = r / m
-
-#         out_f = tempfile.NamedTemporaryFile(delete=False, suffix=".wav")
-#         out_path = out_f.name
-#         out_f.close()
-#         sf.write(out_path, r, samplerate=24000)
-
-#         status = (
-#             "OK\n"
-#             f"speaker: {speaker_name}\n"
-#             f"device: {device}"
-#         )
-#         return out_path, status
-
-#     except Exception:
-#         return None, traceback.format_exc()
 
 @torch.inference_mode()
 def synth_one_speaker(speaker_name: str, text_prompt: str,
@@ -228,7 +173,7 @@ with gr.Blocks() as demo:
     )
 if __name__ == "__main__":
     try:
-        demo.queue(concurrency_count=1, max_size=8).launch(server_name="0.0.0.0", server_port=7860, ssr_mode=False)
+        demo.queue().launch(server_name="0.0.0.0", server_port=7860, ssr_mode=False)
     except TypeError:
         demo.launch(server_name="0.0.0.0", server_port=7860, ssr_mode=False)