English | 简体中文

Video2Notes turns long videos into readable notes. It downloads or imports media, transcribes speech with Whisper, and generates Markdown analysis reports with OpenAI-compatible LLM APIs.

The project is designed for a local workflow: run the UI from the repository root, keep generated files under data/, and reuse cached audio, subtitles, and summaries when you rerun the same video.

• Local Streamlit UI launched with video2notes-ui
• Bilibili, YouTube, and other yt-dlp supported video sources
• Browser cookies.txt support for sites that require login cookies
• Local audio or video import to skip online downloading
• Whisper transcription with auto, cpu, and cuda device modes
• Timestamped subtitle output, with srt as the default pipeline format
• OpenAI-compatible LLM analysis for Chinese or English notes
• Provider presets for OpenAI, SiliconFlow, DeepSeek, Qwen, OpenRouter, Ollama, and LM Studio
• Batch processing from pasted links or a relative URL file
• Relative-path artifact layout under data/audio, data/subtitles, data/summaries, and data/metadata
• Markdown report download from the UI

Clone the repository and create a conda environment:

git clone https://github.com/jackwu925/video2notes.git
cd video2notes
conda create -n video2notes python=3.10 -y
conda activate video2notes
conda install -c conda-forge ffmpeg -y
python -m pip install -U pip

Install PyTorch before installing Video2Notes dependencies. Choose the PyTorch command that matches your operating system, package manager, CUDA version, and GPU driver from the official PyTorch install selector:

pip3 install torch torchvision --index-url https://download.pytorch.org/whl/cu126

After PyTorch is installed, install Video2Notes and its pinned runtime dependencies:

python -m pip install -r requirements.txt

Run all commands from the project root. requirements.txt installs the local checkout, so after activating conda you can start the web UI with:

video2notes-ui

Open the Streamlit URL printed in the terminal, configure your provider in API Configuration, paste a video link, and click Run.

For CLI-only usage without the web UI:

python -m pip install -e .
video2notes run "VIDEO_URL" --reuse-existing

For Bilibili video analysis, export your own browser cookies first. In Chrome, use the Get cookies.txt LOCALLY extension to save cookies as a relative file such as cookies.txt, then set that file in the UI cookies.txt field. Cookie files and generated data are ignored by git.

Generate subtitles only:

video2notes run "VIDEO_URL" --skip-llm

Use a local media file and skip downloading:

video2notes run --media-file data/imports/demo.mp4 --reuse-existing

Process multiple links from a relative file:

video2notes run --url-file data/urls.txt --reuse-existing

CLI LLM configuration can be provided with a local .env file. Create .env in the project root with your OpenAI-compatible provider settings:

LLM_API_KEY=your_api_key
LLM_BASE_URL=https://api.openai.com/v1
LLM_MODEL=gpt-4o-mini
LLM_TIMEOUT_SECONDS=60

Video2Notes is released under the MIT License.