移除 GPT-SoVITS 和 Fish Speech，整理 audio_1 脚本与文档

- 删除 03_gen_tts_gptsovits.py、00_transcribe_refs.py、ref_texts.json - 删除 02_gen_tts_fishspeech.py 及 tts/v2_fishspeech/ 生成结果 - 保留 VoxCPM2 / ACE-Step / Stable Audio 三条链路 - 更新执行文档和技术报告，清除两款 TTS 的所有引用 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-20 23:12:01 +08:00
parent 10ced5cb69
commit e5d3d44f9d
501 changed files with 960 additions and 37 deletions
@@ -0,0 +1,83 @@
+#!/usr/bin/env python3
+"""
+步骤 6：格式整理 — 将 audio_1 目录下所有残留 WAV 转换为 MP3，并清理中间文件
+输出：各子目录内的 *_v1.mp3
+
+使用方法：
+  python audio_1/scripts/06_convert_all_to_mp3.py [--dry-run] [--clean-wav]
+"""
+import os, subprocess, argparse
+
+AUDIO_1_DIR = "/home/xsl/blind/audio_1"
+
+def wav_to_mp3(src, dst):
+    subprocess.run(["ffmpeg", "-y", "-i", src, "-codec:a", "libmp3lame",
+                    "-b:a", "128k", "-ar", "44100", "-ac", "2", dst],
+                   check=True, capture_output=True)
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--dry-run", action="store_true", help="只列出，不执行")
+    parser.add_argument("--clean-wav", action="store_true", help="转换后删除 WAV 源文件")
+    args = parser.parse_args()
+
+    wavs_found = []
+    for root, dirs, files in os.walk(AUDIO_1_DIR):
+        # 跳过 scripts 目录
+        if "scripts" in root:
+            continue
+        for fname in files:
+            if fname.endswith(".wav"):
+                wav_path = os.path.join(root, fname)
+                mp3_path = wav_path.replace(".wav", ".mp3")
+                wavs_found.append((wav_path, mp3_path))
+
+    if not wavs_found:
+        print("无需转换的 WAV 文件。")
+        return
+
+    print(f"\n找到 {len(wavs_found)} 个 WAV 文件\n")
+    ok = skip = fail = 0
+    for wav, mp3 in wavs_found:
+        rel = os.path.relpath(wav, AUDIO_1_DIR)
+        if os.path.exists(mp3):
+            print(f"  [SKIP] {rel}")
+            skip += 1
+            continue
+        if args.dry_run:
+            print(f"  [DRY]  {rel} → {os.path.basename(mp3)}")
+            continue
+        try:
+            wav_to_mp3(wav, mp3)
+            size = os.path.getsize(mp3) // 1024
+            print(f"  [OK]   {rel} → {size}KB")
+            if args.clean_wav:
+                os.remove(wav)
+            ok += 1
+        except Exception as e:
+            print(f"  [FAIL] {rel}: {e}")
+            fail += 1
+
+    print(f"\n完成: {ok} ok, {skip} skip, {fail} fail")
+
+    # 统计各目录 MP3 数量
+    print("\n─── 音频资源统计 ───")
+    subdirs = [
+        ("tts/v1_voxcpm2",        "TTS v1 VoxCPM2"),
+        ("music/v1_acestep/music",    "Music v1 ACE-Step BGM"),
+        ("music/v1_acestep/sfx",      "Music v1 ACE-Step SFX"),
+        ("music/v1_acestep/ambience", "Music v1 ACE-Step Amb"),
+        ("music/v2_stableaudio/music",    "Music v2 StableAudio BGM"),
+        ("music/v2_stableaudio/sfx",      "Music v2 StableAudio SFX"),
+        ("music/v2_stableaudio/ambience", "Music v2 StableAudio Amb"),
+    ]
+    for rel_dir, label in subdirs:
+        full = os.path.join(AUDIO_1_DIR, rel_dir)
+        if not os.path.isdir(full):
+            count = 0
+        else:
+            count = len([f for f in os.listdir(full) if f.endswith(".mp3")])
+        print(f"  {label:<30} {count:>4} 条")
+
+if __name__ == "__main__":
+    main()