## 相關資料

- 動畫瘋匯出（？
- Chrome extension: [Video DownloadHelper](https://chromewebstore.google.com/detail/video-downloadhelper/lmjnegcaeklhafolokijcfjliaokphfk)
- [CoApp](https://www.downloadhelper.net/install-coapp-v2)
- OCR
- [EasyOCR](https://github.com/JaidedAI/EasyOCR)
- [video-subtitle-extractor](https://github.com/YaoFANGUK/video-subtitle-extractor)

> Help and generated by gpt-4o

## MVP Backend System Requirements Specification (SRS)

**Project Name**: Anime Subtitle Search and Meme Frame Generator MVP
**Technology Stack**: Node.js, Express, SQLite, FFmpeg
**Primary Use Case**: Build a backend service that supports video subtitle search and frame extraction for creating memes.

---

## 1. Overview

The MVP backend system will enable the following core functionalities:
- Accept an anime video and a subtitle file (SRT).
- Extract frames from the video based on subtitle timings using FFmpeg.
- Provide a search interface to retrieve subtitle blocks and their corresponding frames.
- Allow users to vote on the most suitable frames for meme creation.
- Store all generated frames, subtitle data, and voting information in a structured and efficient manner.

---

## 2. Functional Requirements

### 2.1 Video and Subtitle Management

**FR-1.1 Video Upload and Storage**
- The system shall accept an anime video file (e.g., MP4) uploaded by the user or obtained from a source (e.g., YouTube).
- The video metadata (e.g., title, duration, file path) shall be stored in the `videos` table in the database.
- The uploaded video file shall be saved on the server's filesystem in a predefined folder structure.

**FR-1.2 Subtitle File Upload and Parsing**
- The system shall accept an SRT subtitle file associated with the uploaded video.
- The subtitle file shall be parsed, and each subtitle block (with start/end times and text) shall be stored in the `subtitles` table in the database, linked to the corresponding video.

---

### 2.2 Frame Extraction and Storage

**FR-2.1 Subtitle to Frame Extraction**
- The system shall use **FFmpeg** to extract frames from the video based on the timing of the subtitle blocks.
- The system shall generate frames at the midpoint of each subtitle block and ±1 second variations.
- The system shall store the extracted frames as PNG files in a structured directory format, organized by video and subtitle.

**FR-2.2 Frame Storage in Database**
- The system shall store the metadata of each generated frame (e.g., video ID, subtitle ID, frame time, file path) in the `frames` table.
- The system shall ensure the generated frames are accessible and retrievable based on subtitle search queries.

---

### 2.3 Subtitle Search and Frame Retrieval

**FR-3.1 Subtitle Search**
- The system shall provide an API that allows users to search for subtitles using keywords.
- The search results shall return subtitle blocks, including start/end times, subtitle text, and corresponding frames.

**FR-3.2 Frame Retrieval**
- For each searched subtitle block, the system shall return all associated frames (including variations) for display in the user interface.
- The system shall allow users to retrieve the file path of a specific frame to download or share the image.

---

### 2.4 Voting on Frames

**FR-4.1 Voting API**
- The system shall provide an API that allows users to upvote or downvote frames.
- Each vote shall be linked to a specific frame in the `votes` table, recording the user (or anonymous ID), vote type (upvote or downvote), and timestamp.

**FR-4.2 Frame Ranking**
- The system shall calculate the ranking of frames based on user votes and return the most popular frame for a specific subtitle block when requested.

---

## 3. Non-Functional Requirements

### 3.1 Performance
- The system shall process and generate frames in under 5 seconds for a 2-minute video clip.
- The search queries for subtitles shall return results in less than 1 second for keyword-based searches.

### 3.2 Scalability
- The system shall be designed to handle multiple video uploads and frame extractions concurrently.

### 3.3 Data Storage
- The system shall use **SQLite** for local development and testing to store video, subtitle, frame, and vote data. **PostgreSQL** may be used for future scalability.
- The extracted frames shall be stored in the filesystem under `/frames/{video_id}/{subtitle_id}/`.

### 3.4 Error Handling
- The system shall provide detailed error responses if the subtitle or video files are not valid or fail to process.
- The system shall handle API request errors with appropriate HTTP status codes (e.g., 400 Bad Request, 500 Internal Server Error).

---

## 4. API Specifications

### 4.1 Upload Video
- **Method**: `POST /api/videos`
- **Input**:
- Video file (MP4, AVI, etc.)
- Video metadata (title, series, episode number)
- **Output**: Video ID, status (success/failure)

### 4.2 Upload Subtitle
- **Method**: `POST /api/subtitles`
- **Input**:
- Subtitle file (SRT)
- Associated video ID
- **Output**: Subtitle processing status (success/failure)

### 4.3 Search Subtitles
- **Method**: `GET /api/subtitles/search`
- **Input**:
- Query string (keywords)
- **Output**: List of subtitle blocks (start time, end time, text) and associated frames

### 4.4 Retrieve Frames for Subtitle
- **Method**: `GET /api/frames/{subtitle_id}`
- **Input**:
- Subtitle ID
- **Output**: List of frames (file paths, time stamps)

### 4.5 Vote on Frame
- **Method**: `POST /api/frames/{frame_id}/vote`
- **Input**:
- Frame ID, vote type (upvote/downvote)
- **Output**: Vote status (success/failure)

---

## 5. Data Models

### 5.1 Video Model

{
  "id": "xyz456",
  "title": "Anime Episode 1",
  "file_path": "/videos/xyz456.mp4",
  "duration": 1440,
  "created_at": "2024-10-12T12:00:00Z",
  "source_type": "user_upload"
}

### 5.2 Subtitle Model

{
  "id": "abc123",
  "video_id": "xyz456",
  "start_time": 120.0,
  "end_time": 125.0,
  "text": "This is a sample subtitle."
}

### 5.3 Frame Model

{
  "id": "def789",
  "video_id": "xyz456",
  "subtitle_id": "abc123",
  "frame_time": 121.0,
  "file_path": "/frames/xyz456/abc123/frame_121000.png"
}

### 5.4 Vote Model

{
  "id": "ghi101",
  "frame_id": "def789",
  "user_id": "user123",
  "vote_type": "upvote",
  "created_at": "2024-10-12T12:05:00Z"
}

### 6. Assumptions and Constraints

* The MVP will initially focus on one specific anime series where subtitle files (SRT) are available.
* The video file sizes should be limited to 500MB for processing efficiency in the MVP.
* The extracted frames will be stored on the server filesystem, and further storage optimizations (like CDN integration) can be added in later versions.