This project provides an automated Bilibili live stream recording and content creation pipeline, targeting users who want to capture, process, and re-upload live content with minimal manual intervention. It aims to simplify the workflow from recording to uploading, including advanced features like automatic subtitle generation and clip creation, even on low-configuration hardware.

How It Works

The engine utilizes a pipeline architecture for efficient processing, aiming for near real-time recording and uploading. Key features include automatic danmaku (bullet comment) conversion and rendering into videos, speech-to-text via OpenAI's Whisper for subtitle generation, and AI-powered automatic clip creation based on danmaku density. It also supports AI-generated video covers and automatic uploading of both full recordings and clips to Bilibili, with options for multi-platform live streaming.

Quick Start & Requirements

• Installation: Clone the repository with submodules (git clone --recurse-submodules) and install dependencies via pip install -r requirements.txt. Ensure FFmpeg is installed.
• Prerequisites: Python 3.10+, FFmpeg. For subtitle generation using local Whisper, an NVIDIA GPU with >2.7GB VRAM and CUDA is required. API-based Whisper and MLLM models require API keys.
• Setup: Configuration is done via bilive.toml and settings.toml. Docker images are available for both CPU and GPU (amd64 only) setups.
• Documentation: English Documentation

Highlighted Details

• Extremely fast recording and processing, claimed to be the fastest stable Bilibili recorder.
• Supports both x64 and arm64 architectures, with minimal hardware requirements (single-core CPU, low RAM).
• Integrates multiple MLLMs (GLM-4V-PLUS, Gemini-2.0-flash, Qwen-2.5-72B-Instruct, SenseNova V6 Pro) for clip title generation and various image generation models for video covers.
• Includes a persistent login tool (bilitool) for seamless uploading and supports multi-part video uploads.

Maintenance & Community

The project actively develops and integrates new AI models. Links to issue reporting and a chat group (via image) are provided.

Licensing & Compatibility

The project is released under the MIT License. It is designed for personal learning and exchange; unauthorized commercial use or large-scale recording is discouraged due to potential platform restrictions.

Limitations & Caveats

The arm64 version of the Docker image does not support local Whisper deployment due to Triton library compatibility issues. Users must ensure sufficient GPU VRAM for local Whisper deployment. API-based features are subject to third-party rate limits and costs.