warren/rust-scribe
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: Add per-segment language detection and dual output files

Browse Source 
- Add whatlang for real-time language detection per segment
- Generate .asr.json (basic) and .asrx.json (with language labels)
- Add auto-save progress with configurable interval
- Add resume functionality for interrupted transcriptions
- Add Music/Empty detection and statistics
- Update progress display with unified format
- Add comprehensive Chinese README with usage documentation
This commit is contained in:
Warren Lo
2026-03-11 18:17:55 +08:00
parent f9653eef58
commit d6af9546ea
4 changed files with 1022 additions and 175 deletions
Download Patch File Download Diff File Expand all files Collapse all files

123
Cargo.lock generated
View File

@@ -2,6 +2,18 @@
# It is not intended for manual editing.
version = 4
[[package]]
name = "ahash"
version = "0.8.12"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5a15f179cd60c4584b8a8c596927aadc462e27f2ca70c04e0071964a73ba7a75"
dependencies = [
"cfg-if",
"once_cell",
"version_check",
"zerocopy",
]
[[package]]
name = "aho-corasick"
version = "1.1.4"
@@ -11,6 +23,12 @@ dependencies = [
"memchr",
]
[[package]]
name = "allocator-api2"
version = "0.2.21"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "683d7910e743518b0e34f1186f92494becacb047c7b6bf616c96772180fef923"
[[package]]
name = "anstream"
version = "0.6.21"
@@ -270,6 +288,16 @@ version = "0.3.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0cc23270f6e1808e30a928bdc84dea0b9b4136a8bc82338574f23baf47bbd280"
[[package]]
name = "hashbrown"
version = "0.14.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e5274423e17b7c9fc20b6e7e208532f9b19825d82dfd615708b70edd83df41f1"
dependencies = [
"ahash",
"allocator-api2",
]
[[package]]
name = "heck"
version = "0.5.0"
@@ -315,6 +343,12 @@ dependencies = [
"either",
]
[[package]]
name = "itoa"
version = "1.0.17"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "92ecc6618181def0457392ccd0ee51198e065e016d1d527a7ac1b6dc7c1f09d2"
[[package]]
name = "lazy_static"
version = "1.5.0"
@@ -525,7 +559,11 @@ dependencies = [
"anyhow",
"clap",
"ffmpeg-next",
"lazy_static",
"ndarray",
"serde",
"serde_json",
"whatlang",
"whisper-rs",
"whisper-rs-sys",
]
@@ -555,6 +593,49 @@ dependencies = [
"windows-sys 0.59.0",
]
[[package]]
name = "serde"
version = "1.0.228"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9a8e94ea7f378bd32cbbd37198a4a91436180c5bb472411e48b5ec2e2124ae9e"
dependencies = [
"serde_core",
"serde_derive",
]
[[package]]
name = "serde_core"
version = "1.0.228"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "41d385c7d4ca58e59fc732af25c3983b67ac852c1a25000afe1175de458b67ad"
dependencies = [
"serde_derive",
]
[[package]]
name = "serde_derive"
version = "1.0.228"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d540f220d3187173da220f885ab66608367b6574e925011a9353e4badda91d79"
dependencies = [
"proc-macro2",
"quote",
"syn",
]
[[package]]
name = "serde_json"
version = "1.0.149"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "83fc039473c5595ace860d8c4fafa220ff474b3fc6bfdb4293327f1a37e94d86"
dependencies = [
"itoa",
"memchr",
"serde",
"serde_core",
"zmij",
]
[[package]]
name = "shlex"
version = "1.3.0"
@@ -596,6 +677,22 @@ version = "0.2.15"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "accd4ea62f7bb7a82fe23066fb0957d48ef677f6eeb8215f372f52e48bb32426"
[[package]]
name = "version_check"
version = "0.9.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0b928f33d975fc6ad9f86c8f283853ad26bdd5b10b7f1542aa2fa15e2289105a"
[[package]]
name = "whatlang"
version = "0.16.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "471d1c1645d361eb782a1650b1786a8fb58dd625e681a04c09f5ff7c8764a7b0"
dependencies = [
"hashbrown",
"once_cell",
]
[[package]]
name = "which"
version = "4.4.2"
@@ -716,3 +813,29 @@ name = "windows_x86_64_msvc"
version = "0.52.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "589f6da84c646204747d1270a2a5661ea66ed1cced2631d546fdfb155959f9ec"
[[package]]
name = "zerocopy"
version = "0.8.42"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f2578b716f8a7a858b7f02d5bd870c14bf4ddbbcf3a4c05414ba6503640505e3"
dependencies = [
"zerocopy-derive",
]
[[package]]
name = "zerocopy-derive"
version = "0.8.42"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7e6cc098ea4d3bd6246687de65af3f920c430e236bee1e3bf2e441463f08a02f"
dependencies = [
"proc-macro2",
"quote",
"syn",
]
[[package]]
name = "zmij"
version = "1.0.21"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b8848ee67ecc8aedbaf3e4122217aff892639231befc6a1b58d29fff4c2cabaa"

4
Cargo.toml
View File

@@ -22,6 +22,10 @@ anyhow = "1.0"
# 可選：如果 whisper-rs 需要額外的數學運算支持
ndarray = "0.15"
serde = { version = "1.0", features = ["derive"] }
serde_json = "1.0"
lazy_static = "1.4"
whatlang = "0.16"

161
README.md
View File

@@ -4,21 +4,25 @@
## 功能特性
- **高效音頻提取**：利用 FFmpeg 直接從影片或音頻檔案中提取並重採樣為 Whisper 所需的格式 (16kHz, Mono, f32)。
- **精準轉寫**：基於 OpenAI Whisper 模型，提供高精度的語音轉文字功能。
- **時間碼支援**：自動產生精確到毫秒的逐字稿時間戳。
- **語言自動檢測**：若未指定語言，可自動偵測輸入音頻的語言。
- **優化進度顯示**：在轉寫過程中提供即時的處理進度、耗時與預估剩餘時間。
- **高效音頻提取**：利用 FFmpeg 直接從影片或音頻檔案中提取並重採樣為 Whisper 所需的格式 (16kHz, Mono, f32)
- **精準轉寫**：基於 OpenAI Whisper 模型，提供高精度的語音轉文字功能
- **時間碼支援**：自動產生精確到毫秒的逐字稿時間戳，格式為 `HH:MM:SS.mmm(fXX)` (25fps)
- **語言自動檢測**：若未指定語言，可自動偵測輸入音頻的語言
- **優化進度顯示**：在轉寫過程中提供即時的處理進度、耗時與預估剩餘時間
- **自動存檔**：支援定期自動存檔，避免長時間轉寫因意外中斷而遺失進度
- **中斷續處理**：支援從上次中斷點繼續處理
- **語言分段檢測**：自動檢測每段文字的語言類型，生成帶語言標籤的逐字稿
- **特殊內容識別**：自動識別 Music (音樂) 和 Empty (空文字) 分段
## 前置需求
在編譯與執行前，請確保系統已安裝以下環境：
1. **Rust 環境**：請安裝 [Rustup](https://rustup.rs/)。
2. **FFmpeg**：系統必須安裝 FFmpeg 開發庫。
1. **Rust 環境**：請安裝 [Rustup](https://rustup.rs/)
2. **FFmpeg**：系統必須安裝 FFmpeg 開發庫
- macOS: `brew install ffmpeg`
- Ubuntu/Debian: `sudo apt install libavcodec-dev libavformat-dev libavutil-dev libswresample-dev`
3. **Whisper 模型**：請準備 Whisper 的 `.bin` 模型檔案 (例如 `ggml-base.bin`)，並放置於 `models/` 目錄中。
3. **Whisper 模型**：請準備 Whisper 的 `.bin` 模型檔案，並放置於 `models/` 目錄中
## 編譯與安裝
@@ -33,7 +37,7 @@ cargo build --release
## 使用方式
執行程式時，需指定輸入檔案路徑及模型路徑：
### 基本使用
```bash
./target/release/rust-scribe <input_file> --model models/<your-model.bin>
@@ -41,33 +45,148 @@ cargo build --release
### 參數說明
- `<input_file>`: 欲轉寫的影片或音頻檔案路徑 (位置參數)。
- `-m, --model <MODEL_PATH>`: Whisper 模型檔案路徑 (必填)。
- `-l, --language <LANG>`: 指定轉寫語言 (例如 `zh`, `en`)。若不指定，系統將自動偵測 (選填)。
- `-v, --verbose`: 開啟詳細轉寫進度輸出 (選填)。
| 參數 | 說明 | 必填 |
|------|------|------|
| `<input_file>` | 欲轉寫的影片或音頻檔案路徑 | ✅ |
| `-m, --model <MODEL_PATH>` | Whisper 模型檔案路徑 | ✅ |
| `-l, --language <LANG>` | 指定轉寫語言 (例如 `zh`, `en`)。若不指定將自動偵測 | ❌ |
| `-v, --verbose` | 開啟詳細轉寫進度輸出 | ❌ |
| `--save-interval <N>` | 每 N 個分段自動存檔一次 (預設: 100) | ❌ |
| `--max-duration <秒>` | 限制處理的最大時長（用於測試） | ❌ |
**使用範例**：
### 使用範例
```bash
# 指定使用 base 模型，自動檢測語言
cargo run --release -- video.mp4 --model models/ggml-base.bin
# 基本使用，自動檢測語言，每 10 個分段存檔
cargo run --release -- video.mp4 --model models/ggml-base.bin --save-interval 10
# 指定使用中文進行轉寫
# 指定中文轉寫
cargo run --release -- lecture.mkv --model models/ggml-base.bin --language zh
# 限制處理前 60 秒（用於測試）
cargo run --release -- video.mp4 --model models/ggml-base.bin --max-duration 60
```
## 輸出檔案
程式會生成兩個輸出檔案：
### 1. `.asr.json` - 基本逐字稿
包含基本的轉寫結果：
```json
{
"input_file": "video.mp4",
"language": "English",
"segments": [
{
"start": "00:00:05.120(f03)",
"end": "00:00:08.560(f14)",
"text": "Hello everyone",
"language": "English"
}
]
}
```
### 2. `.asrx.json` - 帶語言檢測的逐字稿
包含每段文字的語言檢測結果：
```json
{
"input_file": "video.mp4",
"segments": [
{
"start": "00:00:05.120(f03)",
"end": "00:00:08.560(f14)",
"text": "Hello everyone",
"language": "English",
"confidence": 0.98
},
{
"start": "00:00:10.000(f00)",
"end": "00:00:13.000(f00)",
"text": "[Music]",
"language": "Music",
"confidence": 1.0
}
]
}
```
## 進度顯示說明
轉寫過程中會顯示即時進度：
```
═══════════════════════════════════════════════════════════
🎙️ Transcribing (this may take a few minutes)...
💡 Auto-saving every 10 segments
📋 Legend:
P = Processed (已處理分段數)
A = ASR saved (.asr.json 已存檔數)
X = Lang detect saved (.asrx.json 已存檔數)
═══════════════════════════════════════════════════════════
🔄 Progress: [71:05/114:39] 62.0% | Elapsed: 03:04 | Remaining: 01:54 | P:1112 A:1110 X:1110 | English:500 Unknown:52 Music:58
```
### 進度列格式說明
- **時間進度**：`[已處理時間/總時長] 完成百分比`
- **時間統計**：`Elapsed: 已耗時 | Remaining: 預估剩餘時間`
- **分段統計**：`P:已處理數 A:ASR存檔數 X:語言檢測存檔數`
- **語言統計**：各語言類型的分段數量（依序顯示）
### 特殊類型標籤
- **Music**：音樂或背景音效段落
- **Empty**：空文字或純空白段落
- **Unknown**：語言識別置信度低於 0.5 的段落
## 中斷續處理
程式支援中斷後繼續處理：
1. 當程式偵測到已存在的輸出檔案時，會詢問是否繼續
2. 選擇 `[C] Continue` 會從上次中斷點繼續
3. 選擇 `[R] Restart` 會刪除舊檔案重新開始
兩個檔案（`.asr.json` 和 `.asrx.json`）會同步處理。
## 語言檢測說明
使用 [whatlang](https://github.com/greyblake/whatlang-rs) 函式庫進行語言檢測：
- **置信度 ≥ 0.5**：顯示檢測到的語言名稱
- **置信度 < 0.5**：標註為 "Unknown"
- **Music**：自動識別包含 [Music] 或 (music) 標記的段落
- **Empty**：空文字或純空白段落
## 專案結構
```
rust-scribe/
├── src/
│ └── main.rs # 核心邏輯 (音頻處理、Whisper 轉寫、CLI 介面)
├── models/ # 存放 Whisper 模型檔案
├── Cargo.toml # 專案依賴與配置
│ └── main.rs # 核心邏輯
├── models/ # Whisper 模型檔案
├── Cargo.toml # 專案配置
├── Cargo.lock # 依賴鎖定
├── README.md # 使用文檔
├── AGENTS.md # 開發者規範
└── .cargo/
└── config.toml # Cargo 編譯配置
└── config.toml # Cargo 配置
```
## 開發者規範
請參閱 `AGENTS.md` 了解詳細的代碼風格、編譯與測試指南。
## 許可證
MIT License
## 致謝
- [Whisper.cpp](https://github.com/ggerganov/whisper.cpp) - OpenAI Whisper 的 C++ 實現
- [FFmpeg](https://ffmpeg.org/) - 音頻處理函式庫
- [whisper-rs](https://github.com/tazz4843/whisper-rs) - Whisper.cpp 的 Rust 綁定
- [whatlang](https://github.com/greyblake/whatlang-rs) - 自然語言檢測函式庫

873
src/main.rs
View File

File diff suppressed because it is too large Load Diff