Add local GPU preset voices with dropdown selection.

Generate ChatTTS sample_random_speaker presets without cloud APIs; choose clone or preset in synthesize UI.

Co-authored-by: Cursor <cursoragent@cursor.com>

scripts/generate_voice_presets.py

@@ -0,0 +1,80 @@
+#!/usr/bin/env python3
+"""生成 ChatTTS 本地预设说话人（sample_random_speaker，走 GPU）。"""
+
+from __future__ import annotations
+
+import json
+import sys
+from datetime import datetime
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parent.parent
+sys.path.insert(0, str(ROOT))
+
+import torch
+
+from config import CHATTTS_MODEL_DIR
+from tts_service import get_chattts_instance, reset_chattts_instance
+from voice_presets import (
+    DEFAULT_MANIFEST,
+    MANIFEST_PATH,
+    PRESETS_DIR,
+    VOICES_DIR,
+    ensure_manifest,
+)
+
+
+def main() -> None:
+    ensure_manifest()
+    PRESETS_DIR.mkdir(parents=True, exist_ok=True)
+
+    try:
+        from whisper_service import reset_whisper_model
+
+        reset_whisper_model()
+    except Exception:
+        pass
+
+    reset_chattts_instance()
+    chat, err = get_chattts_instance()
+    if chat is None:
+        raise SystemExit(f"ChatTTS 加载失败: {err}")
+
+    if not hasattr(chat, "sample_random_speaker"):
+        raise SystemExit("当前 ChatTTS 版本不支持 sample_random_speaker")
+
+    presets = DEFAULT_MANIFEST["presets"]
+    print(f"[INFO] 生成 {len(presets)} 个预设音色 → {PRESETS_DIR}")
+
+    for item in presets:
+        pid = item["id"]
+        label = item["label"]
+        out_path = PRESETS_DIR / f"{pid}.pt"
+
+        spk_emb = chat.sample_random_speaker()
+        payload = {
+            "version": 1,
+            "preset": True,
+            "id": pid,
+            "label": label,
+            "spk_emb": spk_emb,
+            "spk_smp": None,
+            "txt_smp": "",
+            "created_at": datetime.now().isoformat(),
+            "source": "ChatTTS.sample_random_speaker",
+        }
+        torch.save(payload, out_path)
+        print(f"  [OK] {label} → {out_path.name}")
+
+    manifest = json.loads(MANIFEST_PATH.read_text(encoding="utf-8"))
+    manifest["generated_at"] = datetime.now().isoformat()
+    manifest["chattts_model"] = str(CHATTTS_MODEL_DIR)
+    MANIFEST_PATH.write_text(
+        json.dumps(manifest, ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+    print("[OK] 全部预设音色生成完成")
+
+
+if __name__ == "__main__":
+    main()

scripts/generate_voice_presets.sh

@@ -0,0 +1,16 @@
+#!/usr/bin/env bash
+# 生成本地 GPU 预设音色（ChatTTS 内置说话人，无需 API）
+# 用法: bash scripts/generate_voice_presets.sh
+set -euo pipefail
+ROOT="$(cd "$(dirname "$0")/.." && pwd)"
+VENV_PY="${ROOT}/venv/bin/python"
+
+if [[ ! -x "${VENV_PY}" ]]; then
+  echo "[ERROR] 未找到 venv，请先 bash deploy.sh deps"
+  exit 1
+fi
+
+echo "[INFO] 正在生成 ChatTTS 预设音色（本地 GPU）..."
+"${VENV_PY}" "${ROOT}/scripts/generate_voice_presets.py"
+echo "[OK] 预设音色已写入 ${ROOT}/voices/presets/"
+echo "[OK] 在 Web UI「配音合成」处可从下拉框选择"