• Auto-download course videos from SJTU Canvas as soon as they are available.
• Merge video clips of successive lectures into one file.
• Transcribe accurately based on OpenAI's whisper (GPU recommended).
• Summarize. (TODO)

PRs are welcome!

Important: Linux support only (Currently).

This project is not affiliated with SJTU or Canvas. It is for educational purposes only. Please respect the intellectual property rights of the content creators. The author is not responsible for any misuse of this project.

• Clone the repository.
• Install the required packages using pip install -r requirements.txt.
• Install aria2 and add it to your PATH.
• Install ffmpeg and add it to your PATH.
• (Optional) Install whisper-ctranslate2 to enable transcription. (GPU recommended)
• Write configuration in config.toml (see below).

Default configuration file is ~/.config/forsythia/config.toml. You can also specify the configuration file using the -c option.

Basic example:

username = "username" # your SJTU username
password = "password" # your SJTU password. If not provided, you will be prompted to enter it.

video_dir = "~/Videos/forsythia"

# Important: Only courses added to the list will be auto-downloaded.
[course.MA]	# the first course to auto-download
course_id = 64032
# week : number of videos to download
course_table = { 1 = 2, 3 = 2, 5 = 2 }
# the above line means that for Monday, Wednesday and Friday, download 2 videos each.

[course.PH]	# the second course to auto-download
course_id = 64222
course_table = { 1 = 2, 3 = 2, 5 = 2 }

For more options, see config.py in the repository. You may refer to the TOML documentation and the pydantic documentation for more information.

Auto-download videos:

./forsythia.py auto_download -c /path/to/config.toml

Note: For daemon mode, consider nohup, tmux or systemd.

Download videos for a specific course on a specific day:

./forsythia.py search_download -n PH -m 2 --transcribe --readable_subtitle 64222

Merge (and transcribe) video clips:

$./forsythia.py process_video -h
usage: forsythia.py process_video [-h] -o OUTPUT_FILE [-t TMP_PATH] [--transcribe] [--readable_subtitle] input_files [input_files ...]

Process and merge videos with optional transcription and readable subtitles.

positional arguments:
  input_files           List of input video files or URLs to process.

options:
  -h, --help            show this help message and exit
  -o OUTPUT_FILE, --output_file OUTPUT_FILE
                        Output file path for the processed video.
  -t TMP_PATH, --tmp_path TMP_PATH
                        Temporary directory for processing files. Defaults to a system temp directory.
  --transcribe          Enable video transcription.
  --readable_subtitle   Generate a readable subtitle file (requires --transcribe).

• sjtu-canvas-video-download for SJTU login and download scripts.