ishenwei/atlas
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

manifest.json里extract audio, transcribe audio 共用一条记录

Browse Source
This commit is contained in:
admin
2026-09-17 08:49:16 +08:00
parent 922c3b4071
commit 44110a0ccf
3 changed files with 235 additions and 110 deletions
Download Patch File Download Diff File Expand all files Collapse all files

125
synthmind/README.md
View File

@@ -47,7 +47,7 @@
- 支持指定语言（`--language zh/en/...`）或自动检测
- 支持音频**切分转录**：超长音频切分后逐段转录再合并
- 支持多种音频格式：`.mp3/.m4a/.wav/.flac/.ogg/.aac`
- **Manifest key 含模型维度**：多模型多参数结果共存，互不覆盖
- **Manifest 嵌套 transcriptions**：同一视频/音频记录里，多模型多参数结果按 `<model>[:<suffix>]` 为 subkey 共存，互不覆盖
---
@@ -344,55 +344,78 @@ python3 transcribe_audio.py /home/user/audio/ --retry-failed
## Manifest 文件说明
两个脚本**共享同一个 `manifest.json`**，通过 key 前缀区分：
两个脚本**共享同一个 `manifest.json`**（schema v2）。每个视频文件在 manifest 中**只有一条记录**，`extract_audio.py` 与 `transcribe_audio.py` 分别写入其中的 `extraction` 与 `transcriptions` 两个子对象，互不覆盖。
- **视频提取记录**：key = 视频绝对路径
- **音频转录记录**：key = `transcribe:<model>[:<suffix>]:<音频绝对路径>`
### key 选择规则
### key 结构详解
| Key | 来源 | 场景 |
|-----|------|------|
| 视频绝对路径（如 `/path/to/x.mp4`）| `extract_audio.py` 产生的记录 | 完整流水线（视频 → 音频 → 转录）|
| 音频绝对路径（如 `/path/to/x.mp3`）| 直接对 mp3 跑 `transcribe_audio.py` | 独立音频转录（无对应视频）|
| Key 格式 | 场景 |
|---------|------|
| `/path/to/x.mp4` | 视频提取记录 |
| `transcribe:tiny:/path/to/x.mp3` | tiny 模型转录（无 suffix） |
| `transcribe:medium:/path/to/x.mp3` | medium 模型转录（无 suffix） |
| `transcribe:base:_v1:/path/to/x.mp3` | base 模型 + `_v1` 后缀 |
| `transcribe:base:_v2:/path/to/x.mp3` | base 模型 + `_v2` 后缀 |
**反向查找**：当 `transcribe_audio.py` 处理某个 mp3 时，会先扫描 manifest，如果发现某条视频记录的 `extraction.audio_path` 等于该 mp3 路径，就把转录结果并入该视频记录，而不是新建 key。
多个 key 可并存于同一 manifest，互不覆盖。
### manifest.json 完整示例（含多模型对比）
### 完整示例（v2）
```json
{
"version": 1,
"version": 2,
"files": {
"/path/to/lesson1.mp4": {
"audio_extracted": true,
"audio_path": "/path/to/lesson1.mp3",
"split_segments": [],
"error": null,
"updated_at": "2026-08-23T10:30:00"
"extraction": {
"extracted": true,
"audio_path": "/path/to/lesson1.mp3",
"split_segments": [],
"error": null,
"updated_at": "2026-08-23T10:30:00"
},
"transcriptions": {
"tiny": {
"transcribed": true,
"txt_path": "/path/to/lesson1.tiny.txt",
"split_segments": [],
"model": "tiny",
"language": "zh",
"output_suffix": null,
"error": null,
"updated_at": "2026-08-23T10:35:00"
},
"medium": {
"transcribed": true,
"txt_path": "/path/to/lesson1.medium.txt",
"split_segments": [],
"model": "medium",
"language": "zh",
"output_suffix": null,
"error": null,
"updated_at": "2026-08-23T10:50:00"
},
"base:_v1": {
"transcribed": true,
"txt_path": "/path/to/lesson1_v1.txt",
"split_segments": [],
"model": "base",
"language": null,
"output_suffix": "_v1",
"error": null,
"updated_at": "2026-08-23T11:00:00"
}
}
},
"transcribe:tiny:/path/to/lesson1.mp3": {
"transcribed": true,
"txt_path": "/path/to/lesson1.tiny.txt",
"split_segments": [],
"model": "tiny",
"language": "zh",
"output_suffix": null,
"error": null,
"updated_at": "2026-08-23T10:35:00"
},
"transcribe:medium:/path/to/lesson1.mp3": {
"transcribed": true,
"txt_path": "/path/to/lesson1.medium.txt",
"split_segments": [],
"model": "medium",
"language": "zh",
"output_suffix": null,
"error": null,
"updated_at": "2026-08-23T10:50:00"
"/path/to/standalone.mp3": {
"extraction": null,
"transcriptions": {
"medium": {
"transcribed": true,
"txt_path": "/path/to/standalone.medium.txt",
"split_segments": [],
"model": "medium",
"language": "en",
"output_suffix": null,
"error": null,
"updated_at": "2026-08-23T11:15:00"
}
}
}
}
}
@@ -400,14 +423,18 @@ python3 transcribe_audio.py /home/user/audio/ --retry-failed
### 字段说明
**视频提取记录字段**
- `audio_extracted` - 是否已提取音频（布尔）
**顶层每条记录**
- `extraction` - `extract_audio.py` 的产物；`null` 表示没有走过提取（例如直接对 mp3 跑转录）
- `transcriptions` - `transcribe_audio.py` 的产物；以 `<model>` 或 `<model>:<suffix>` 为 subkey，允许同一音频用多种模型/参数并存
**`extraction` 字段**
- `extracted` - 是否已提取音频（布尔）；`false` + 有 `error` 表示上次失败
- `audio_path` - 生成的 MP3 路径（切分模式下为第一个片段）
- `split_segments` - 切分片段列表（非切分模式为空）
- `split_segments` - 切分片段列表（非切分模式为空 `[]`）
- `error` - 失败时的错误信息，成功为 `null`
- `updated_at` - 最后更新时间（ISO 8601）
**音频转录记录字段**
**`transcriptions.<subkey>` 字段**
- `transcribed` - 是否已转录（布尔）
- `txt_path` - 生成的 TXT 路径
- `split_segments` - 切分片段列表
@@ -420,12 +447,18 @@ python3 transcribe_audio.py /home/user/audio/ --retry-failed
### 断点续传逻辑
脚本判断是否需要处理某文件时，**同时校验**：
1. manifest 中标记为已完成
1. manifest 中标记为已完成（`extraction.extracted=true` 或 `transcriptions.<sk>.transcribed=true`）
2. 目标文件实际存在且非空
**任一条件不满足则重新处理**（例如手动删除了 mp3 后重跑，会重新提取）。
**多模型模式**：每个 (音频 × 模型 × 后缀) 组合独立跟踪。已完成的组合被跳过，未完成的继续。
**多模型模式**：每个 (音频 × 模型 × 后缀) 组合独立跟踪，对应 `transcriptions` 里独立的 subkey。已完成的组合被跳过，未完成的继续。
### 从旧版 (v1) 升级
老版本使用扁平 schema：视频记录 key = 视频路径；转录记录 key = `transcribe:<model>[:<suffix>]:<音频路径>`。这些散落的记录在 v2 里会被合并到同一条视频记录里。
**升级方式**：脚本首次加载旧 manifest 时会打印警告并**以空 v2 schema 重新开始**（旧数据将被下次保存覆盖）。如果你想保留旧记录，运行升级前**先手动备份 manifest.json**。所有已提取的 mp3 和已转录的 txt 会在下次运行时被识别为"存在但 manifest 无记录"→ 重新处理并写入新 schema。
---

94
synthmind/extract_audio.py
View File

@@ -56,13 +56,32 @@ def log(msg: str, level: str = "INFO"):
print(f"[{ts}] {prefix} {msg}", flush=True)
# ─── Manifest 操作 ─────────────────────────────────────────────────────────────
# ─── Manifest 操作 (v2 schema) ────────────────────────────────────────────────
#
# v2 schema：每个条目形如
# {
# "extraction": { "extracted": bool, "audio_path": str, "split_segments": [],
# "error": str|None, "updated_at": str },
# "transcriptions": {
# "<model>[:<suffix>]": { "transcribed": bool, "txt_path": str, ... }
# }
# }
# transcribe_audio.py 会把转录结果并入同一条视频记录的 transcriptions 里。
# ─────────────────────────────────────────────────────────────────────────────
MANIFEST_VERSION = 2
def load_manifest(manifest_path: Path) -> dict:
if manifest_path.exists():
with open(manifest_path, encoding="utf-8") as f:
return json.load(f)
return {"version": 1, "files": {}}
if not manifest_path.exists():
return {"version": MANIFEST_VERSION, "files": {}}
with open(manifest_path, encoding="utf-8") as f:
data = json.load(f)
# 旧版 manifest (v1) 直接废弃，从头开始（用户已明确接受）
if data.get("version", 1) < MANIFEST_VERSION:
log("检测到旧版 manifest.json (v1)，将以 v2 schema 重新初始化 - 旧记录将被覆盖", "WARN")
return {"version": MANIFEST_VERSION, "files": {}}
return data
def save_manifest(manifest_path: Path, manifest: dict):
@@ -72,34 +91,41 @@ def save_manifest(manifest_path: Path, manifest: dict):
def get_file_entry(manifest: dict, video_path: str) -> dict:
"""获取或初始化某个视频文件的 manifest 条目"""
"""获取或初始化某个视频文件的 manifest 条目 (v2 schema)"""
if video_path not in manifest["files"]:
manifest["files"][video_path] = {
"audio_extracted": False,
"audio_path": None,
"split_segments": [], # 切分后的片段列表（若有）
"error": None,
"updated_at": None,
"extraction": None,
"transcriptions": {},
}
return manifest["files"][video_path]
entry = manifest["files"][video_path]
if "transcriptions" not in entry:
entry["transcriptions"] = {}
return entry
def mark_done(manifest: dict, manifest_path: Path, video_path: str,
audio_path: str, segments: list = None):
entry = get_file_entry(manifest, video_path)
entry["audio_extracted"] = True
entry["audio_path"] = audio_path
entry["split_segments"] = segments or []
entry["error"] = None
entry["updated_at"] = time.strftime("%Y-%m-%dT%H:%M:%S")
entry["extraction"] = {
"extracted": True,
"audio_path": audio_path,
"split_segments": segments or [],
"error": None,
"updated_at": time.strftime("%Y-%m-%dT%H:%M:%S"),
}
save_manifest(manifest_path, manifest)
def mark_failed(manifest: dict, manifest_path: Path, video_path: str, error: str):
entry = get_file_entry(manifest, video_path)
entry["audio_extracted"] = False
entry["error"] = error
entry["updated_at"] = time.strftime("%Y-%m-%dT%H:%M:%S")
prev = entry.get("extraction") or {}
entry["extraction"] = {
"extracted": False,
"audio_path": prev.get("audio_path"),
"split_segments": prev.get("split_segments") or [],
"error": error,
"updated_at": time.strftime("%Y-%m-%dT%H:%M:%S"),
}
save_manifest(manifest_path, manifest)
@@ -220,22 +246,25 @@ def show_status(manifest: dict, all_videos: list):
print("\n" + "=" * 60)
print("📊 提取音频进度状态")
print("=" * 60)
done = sum(1 for v in all_videos
if manifest["files"].get(v, {}).get("audio_extracted"))
failed = sum(1 for v in all_videos
if manifest["files"].get(v, {}).get("error"))
pending = len(all_videos) - done
done, failed = 0, 0
for v in all_videos:
ext = ((manifest["files"].get(v) or {}).get("extraction")) or {}
if ext.get("extracted"):
done += 1
elif ext.get("error"):
failed += 1
pending = len(all_videos) - done - failed
print(f" 总文件数 : {len(all_videos)}")
print(f" 已完成 : {done}")
print(f" 失败 : {failed}")
print(f" 待处理 : {pending}")
print("=" * 60)
for v in all_videos:
entry = manifest["files"].get(v, {})
if entry.get("audio_extracted"):
ext = ((manifest["files"].get(v) or {}).get("extraction")) or {}
if ext.get("extracted"):
status = "✅ 完成"
elif entry.get("error"):
status = f"❌ 失败: {entry['error']}"
elif ext.get("error"):
status = f"❌ 失败: {ext['error']}"
else:
status = "⏳ 待处理"
print(f" {status} {Path(v).name}")
@@ -292,9 +321,10 @@ def main():
pending = []
for v in all_videos:
entry = get_file_entry(manifest, v)
already_done = entry.get("audio_extracted") and entry.get("audio_path") and \
mp3_exists(entry["audio_path"])
is_failed = bool(entry.get("error"))
ext = entry.get("extraction") or {}
already_done = ext.get("extracted") and ext.get("audio_path") and \
mp3_exists(ext["audio_path"])
is_failed = bool(ext.get("error"))
if already_done:
continue

126
synthmind/transcribe_audio.py
View File

@@ -66,13 +66,38 @@ def log(msg: str, level: str = "INFO"):
print(f"[{ts}] {prefix} {msg}", flush=True)
# ─── Manifest 操作 (key 包含 model，支持同一音频多模型共存) ────────────────────
# ─── Manifest 操作 (v2 schema) ────────────────────────────────────────────────
#
# v2 schema：每条记录以 视频路径 (extract_audio 流) 或 音频路径 (独立转录) 为 key，
# 包含 extraction + transcriptions 两部分。extract_audio 与 transcribe_audio
# 共写同一条记录，互不覆盖。
#
# {
# "<key>": {
# "extraction": { "extracted": bool, "audio_path": str, "split_segments": [],
# "error": str|None, "updated_at": str } | null,
# "transcriptions": {
# "<model>[:<suffix>]": { "transcribed": bool, "txt_path": str, "model": str,
# "language": str|None, "output_suffix": str|None,
# "split_segments": [], "error": str|None,
# "updated_at": str }
# }
# }
# }
# ─────────────────────────────────────────────────────────────────────────────
MANIFEST_VERSION = 2
def load_manifest(manifest_path: Path) -> dict:
if manifest_path.exists():
with open(manifest_path, encoding="utf-8") as f:
return json.load(f)
return {"version": 1, "files": {}}
if not manifest_path.exists():
return {"version": MANIFEST_VERSION, "files": {}}
with open(manifest_path, encoding="utf-8") as f:
data = json.load(f)
if data.get("version", 1) < MANIFEST_VERSION:
log("检测到旧版 manifest.json (v1)，将以 v2 schema 重新初始化 - 旧记录将被覆盖", "WARN")
return {"version": MANIFEST_VERSION, "files": {}}
return data
def save_manifest(manifest_path: Path, manifest: dict):
@@ -81,31 +106,56 @@ def save_manifest(manifest_path: Path, manifest: dict):
json.dump(manifest, f, ensure_ascii=False, indent=2)
def transcribe_key(model: str, audio_path: str, output_suffix: str = "") -> str:
"""
manifest key 由 model + output_suffix + 音频路径共同决定，
因此改变 --output-suffix 会形成新的 key（不会误跳过）。
"""
def transcription_subkey(model: str, output_suffix: str = "") -> str:
"""transcriptions 字典的 subkey：<model> 或 <model>:<suffix>"""
if output_suffix:
return f"transcribe:{model}:{output_suffix}:{audio_path}"
return f"transcribe:{model}:{audio_path}"
return f"{model}:{output_suffix}"
return model
def find_entry_key_for_audio(manifest: dict, audio_path: str) -> str:
"""
反向查找：audio 属于哪条 manifest 记录。
- 若某视频记录的 extraction.audio_path == audio → 归入该视频 key
- 否则以 audio_path 自身为 key（独立音频场景）
"""
for key, entry in manifest["files"].items():
extraction = entry.get("extraction")
if extraction and extraction.get("audio_path") == audio_path:
return key
return audio_path
def ensure_entry(manifest: dict, key: str) -> dict:
if key not in manifest["files"]:
manifest["files"][key] = {
"extraction": None,
"transcriptions": {},
}
entry = manifest["files"][key]
if "transcriptions" not in entry:
entry["transcriptions"] = {}
return entry
def get_transcribe_entry(manifest: dict, audio_path: str, model: str,
output_suffix: str = "") -> dict:
key = transcribe_key(model, audio_path, output_suffix)
if key not in manifest["files"]:
manifest["files"][key] = {
"""获取或初始化某个 (音频, 模型, suffix) 的转录记录"""
key = find_entry_key_for_audio(manifest, audio_path)
parent = ensure_entry(manifest, key)
subkey = transcription_subkey(model, output_suffix)
if subkey not in parent["transcriptions"]:
parent["transcriptions"][subkey] = {
"transcribed": False,
"txt_path": None,
"split_segments": [],
"error": None,
"model": model,
"language": None,
"output_suffix": output_suffix or None,
"error": None,
"updated_at": None,
}
return manifest["files"][key]
return parent["transcriptions"][subkey]
def mark_transcribed(manifest: dict, manifest_path: Path, audio_path: str,
@@ -305,31 +355,43 @@ def show_status(manifest: dict, all_audios: list):
print("📊 音频转录进度状态（按 模型 x 音频 展示）")
print("=" * 70)
entries = {k: v for k, v in manifest["files"].items() if k.startswith("transcribe:")}
done = sum(1 for v in entries.values() if v.get("transcribed"))
failed = sum(1 for v in entries.values() if v.get("error"))
print(f" Manifest 记录: {len(entries)} 已完成: {done} 失败: {failed}")
all_trans = []
for parent_key, entry in manifest["files"].items():
for subkey, tr in (entry.get("transcriptions") or {}).items():
all_trans.append((parent_key, subkey, tr))
done = sum(1 for _, _, v in all_trans if v.get("transcribed"))
failed = sum(1 for _, _, v in all_trans if v.get("error"))
print(f" 转录记录: {len(all_trans)} 已完成: {done} 失败: {failed}")
print("=" * 70)
for a in all_audios:
print(f"\n 📄 {Path(a).name}")
matching = [(k, v) for k, v in entries.items() if k.endswith(f":{a}")]
matching = []
for parent_key, entry in manifest["files"].items():
extraction = entry.get("extraction") or {}
is_match = extraction.get("audio_path") == a or parent_key == a
if not is_match:
continue
for subkey, tr in (entry.get("transcriptions") or {}).items():
matching.append((subkey, tr))
if not matching:
print(f" ⏳ 无记录（未转录）")
continue
for key, entry in matching:
m = entry.get("model", "?")
suffix = entry.get("output_suffix") or ""
for subkey, tr in matching:
m = tr.get("model", "?")
suffix = tr.get("output_suffix") or ""
suffix_tag = f" suffix={suffix}" if suffix else ""
if entry.get("transcribed"):
if tr.get("transcribed"):
size = ""
if entry.get("txt_path") and os.path.exists(entry["txt_path"]):
size = f" {os.path.getsize(entry['txt_path']) // 1024}KB"
lang = entry.get("language") or "auto"
txt_name = Path(entry['txt_path']).name if entry.get('txt_path') else '?'
if tr.get("txt_path") and os.path.exists(tr["txt_path"]):
size = f" {os.path.getsize(tr['txt_path']) // 1024}KB"
lang = tr.get("language") or "auto"
txt_name = Path(tr['txt_path']).name if tr.get('txt_path') else '?'
print(f" ✅ [{m:6s}] lang={lang}{suffix_tag}{size} → {txt_name}")
elif entry.get("error"):
print(f" ❌ [{m:6s}]{suffix_tag} {entry['error'][:60]}")
elif tr.get("error"):
print(f" ❌ [{m:6s}]{suffix_tag} {tr['error'][:60]}")
print()