feat: 语音流式输入管线 + VAD前端集成 + 插件-工具合并清理

- 前端: VAD语音检测(@ricky0123/vad-web) + useVoiceInput双模式(流式WS/REST) - Gateway: VoiceStreamManager代理WS流式STT到voice-service - Voice-service: DashScope REST → Realtime WS → Whisper三级引擎 + ffmpeg转码 - 共享模块: pkg/audio(音频转换) + pkg/dashscope(ASR REST客户端) - 清理: 移除旧plugin-manager和pkg/plugins，完成插件→工具合并 - 文档: 完善gateway-api.md和voice-service.md语音API文档 - 工具: scripts/voice/ 语音转换脚本集 Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-06-06 11:50:40 +08:00
parent 258cf81b25
commit 6ef9e082a6
91 changed files with 4091 additions and 3929 deletions
@@ -0,0 +1,100 @@
+#!/usr/bin/env python3
+"""
+批量 WEM → WAV 转换，使用 vgmstream-cli + ffmpeg 标准化。
+输出: 22050Hz, mono, 16-bit PCM WAV
+"""
+
+import os
+import subprocess
+import sys
+from pathlib import Path
+
+VGMSTREAM = r"D:\Project\Code\Uni\Cyrene\scripts\voice\tools\vgmstream\vgmstream-cli.exe"
+RAW_DIR = r"D:\Project\Code\Uni\Cyrene-Voice-Model\data\raw"
+CLEANED_DIR = r"D:\Project\Code\Uni\Cyrene-Voice-Model\data\cleaned"
+
+# 要转换的子目录（按优先级）
+TARGETS = [
+    "VoBanks27", "VoBanks28", "VoBanks29", "VoBanks30", "VoBanks31",
+]
+
+
+def convert_wem_to_wav(wem_path: str, wav_path: str) -> bool:
+    """vgmstream WEM→临时WAV, ffmpeg 标准化 → 最终WAV (22050Hz mono s16)"""
+    os.makedirs(os.path.dirname(wav_path), exist_ok=True)
+
+    # 跳过已存在且非空的文件
+    if os.path.exists(wav_path) and os.path.getsize(wav_path) > 100:
+        return True
+
+    tmp_path = wav_path + ".tmp.wav"
+    try:
+        # Step 1: vgmstream → temp WAV
+        result = subprocess.run(
+            [VGMSTREAM, "-o", tmp_path, wem_path],
+            capture_output=True, timeout=30,
+        )
+        if result.returncode != 0 or not os.path.exists(tmp_path):
+            return False
+
+        # Step 2: ffmpeg → 标准化 22050Hz mono s16
+        result = subprocess.run(
+            ["ffmpeg", "-y", "-i", tmp_path,
+             "-ar", "22050", "-ac", "1", "-sample_fmt", "s16",
+             wav_path],
+            capture_output=True, timeout=30,
+        )
+        if result.returncode != 0:
+            return False
+
+        return os.path.exists(wav_path) and os.path.getsize(wav_path) > 100
+
+    except Exception as e:
+        print(f"  FAIL [{os.path.basename(wem_path)}]: {e}")
+        return False
+    finally:
+        # 清理临时文件
+        if os.path.exists(tmp_path):
+            os.remove(tmp_path)
+
+
+def main():
+    print("=== 批量 WEM → WAV 转换 (VoBanks) ===\n")
+
+    total = 0
+    ok = 0
+    fail = 0
+
+    for target in TARGETS:
+        src_dir = os.path.join(RAW_DIR, target)
+        dst_dir = os.path.join(CLEANED_DIR, target)
+
+        if not os.path.isdir(src_dir):
+            print(f"SKIP: {target} (not found)")
+            continue
+
+        wem_files = sorted(Path(src_dir).glob("*.wem"))
+        if not wem_files:
+            print(f"SKIP: {target} (empty)")
+            continue
+
+        print(f"[{target}] {len(wem_files)} files...")
+
+        for i, wem in enumerate(wem_files):
+            wav = os.path.join(dst_dir, wem.stem + ".wav")
+            if convert_wem_to_wav(str(wem), wav):
+                ok += 1
+            else:
+                fail += 1
+            total += 1
+
+            if (i + 1) % 50 == 0:
+                print(f"  {i+1}/{len(wem_files)} (ok:{ok} fail:{fail})")
+
+        print(f"  Done: {len(wem_files)} files\n")
+
+    print(f"=== 转换完成: {ok} ok, {fail} fail, {total} total ===")
+
+
+if __name__ == "__main__":
+    main()