bilive

bilive is a fast, open-source Bilibili live stream recording and automatic processing tool that records, renders danmaku and subtitles, clips highlights, and uploads videos to Bilibili, designed to run on low-end hardware including ARM64 servers.

What is bilive?

bilive is an automated Bilibili live stream recording and post-production system developed by the GitHub user timerring. It monitors multiple Bilibili live rooms 24/7, records the video stream and danmaku (bullet comments), and then processes the recording into a polished final product with embedded danmaku, speech-to-text subtitles, and automatically generated highlights and covers. The tool runs as a service on Linux (including x64 and ARM64), can be deployed via Docker, and requires no GPU for basic operation, making it suitable for low-cost cloud servers or old hardware.

Key Features

• Pipeline processing — Processes video in a streaming pipeline so the upload is ready within about 30 minutes of the live stream ending, one of the fastest stable Bilibili recording setups.
• Multi-room recording — Simultaneously record any number of Bilibili live rooms, each with its own settings, danmaku logs (including paid messages and gifts), and separate output.
• Automatic danmaku rendering — Converts the XML danmaku file to ASS format (using the open-source DanmakuConvert submodule) and burns it into the video, producing a "with-danmaku" version that is automatically uploaded.
• Whisper speech recognition — Uses OpenAI's Whisper model (either via the Groq API or a local GPU deployment) to transcribe spoken audio into subtitles, which are then rendered into the video as a subtitle track.
• Smart highlight clipping — Analyzes danmaku density to find high-energy segments, automatically cuts them into clips, and generates creative titles using a multimodal large language model (supports GLM-4V-PLUS, Gemini-2.5-flash, Qwen-2.5-72B-Instruct, SenseNova V6 Pro).
• AI cover generation — Automatically generates style-transformed video covers from screenshots using image generation models (Minimax image-01, Kolors, Hunyuan, Stable Diffusion 3.5, etc.).
• Low hardware requirements — Tested on a 1-core ARM64 VPS with 4GB RAM and no GPU; can run on decade-old hardware. No GPU needed for core recording and uploading tasks.
• Persistent login and upload — Uses the bilitool submodule for persistent cookie-based login, supports multi-P upload, and auto-deletes local files after successful upload to save disk space.
• Docker support — Official Docker images available for both CPU-only and GPU (CUDA) deployments, with simple compose configuration.

Who is it for?

• Live stream archivists — Anyone who wants to record and preserve their favorite Bilibili streamers' content automatically, even on a cheap VPS with limited resources.
• Content creators — Streamers or editors who want to quickly turn live streams into polished VODs with danmaku and subtitles, and automatically publish highlight clips.
• Fans and communities — Fan groups that maintain recording archives for multiple streamers and need an automated, low-maintenance solution that runs 24/7.

How does bilive work?

bilive splits the process into three modes: pipeline (fastest, requires GPU for parallel ASR and rendering), append (serial ASR then rendering, lower VRAM), and merge (waits for entire stream to finish, outputs a single full recording). The recording is handled by the blrec library with a web UI for live configuration, while uploading uses bilitool. The typical workflow is: clone the repo, install dependencies, configure bilive.toml and settings.toml, run ./record.sh to start recording, then ./upload.sh to begin processing and uploading finished videos.

FAQ

Does bilive require a GPU?

No. The core recording and danmaku rendering work on CPU only. GPU is optional for local Whisper speech recognition (NVIDIA with >2.7GB VRAM) and for the pipeline mode's parallel processing. The default ASR method is set to none to avoid GPU dependency.

Can I record multiple streams at once?

Yes. Add multiple [[tasks]] entries in settings.toml with different room IDs. Each room gets its own recording and processing pipeline.

Does bilive automatically upload to Bilibili?

Yes. After recording and processing, bilive uses the bilitool submodule to upload videos (including multi-P segments) and highlight clips to your Bilibili account. You must log in once via QR code using the Bilibili app.

Is bilive free?

Yes. bilive is open-source under the Apache-2.0 license. All features are free. You only pay for any external API services you choose to use (e.g., Groq Whisper API, MLLM model API keys).

What platforms does bilive support?

Linux (amd64 and arm64). Windows users can run it via WSL. Docker images are available for both architectures.

Use cases

• 24/7 archival recording — Set up bilive on a $5/month VPS to record multiple streamers around the clock, automatically uploading finished recordings to your Bilibili archive channel.
• Automated highlight reels — Use the danmaku density slicing and MLLM title generation to create shareable highlight clips that publish minutes after a stream ends.
• Accessibility-friendly VODs — Enable Whisper subtitle rendering to make live content accessible to deaf or hard-of-hearing viewers, with no manual transcription required.