BiliSub

For the method of obtaining cookies at Station B, see the application or the attached location Core Highlights (Why BiliSub?) 1.💰Extreme cost control strategy Most similar products on the market brainlessly call the speech recognition API, regardless of whether the video comes with subtitles, resulting in high costs and slow speeds. BiliSub adopts an intelligent dynamic dual-mode extraction strategy to automatically select the optimal route based on your configuration: API Key + Cookies are both valid (recommended): Priority extraction of self-contained subtitles in seconds (0 cost); videos without subtitles automatically call Paraformer for voice transcriptions. ˇOnly API Key is valid: Skip subtitle extraction at Station B (Station B security policy, no correct subtitles will be returned without cookies), and directly force Paraformer voice transcriptions on all videos to ensure the success rate of extraction. ˇOnly cookies are valid: Only the subtitles (CC/hard subtitles) that come with Station B are extracted, and videos without subtitles will be directly marked and failed (no API quota is consumed). Cost estimation: Voice transcriptions only cost about 0.3 yuan/hour (video duration), and about 3 hours of high-quality uncaptioned video content can be processed with a recharge of 1 yuan. Video collection extraction: Support one-click extraction of the entire collection and intelligently schedule concurrent tasks. 2.ˇAmazing transcription efficiency Thanks to the powerful performance of Alibaba Cloud's Paraformer-v2 model, the transcription speed far exceeds that of traditional Whisper. Measured data: For a 30-minute video without subtitles, it only takes 1-2 minutes to complete the full text transliteration. 3. Two operating modes (both support high concurrency) Local direct chain mode (recommended): If you have a public IP or domain name (supports HTTP and HTTPS), the system will directly generate a direct link and submit it to Alibaba Cloud Supports 8 concurrent multiple tasks simultaneously extraction, stable and high speed Third-party direct chain model (out of the box)ˇ NEW: Intelligent concurrency detection: simultaneously detects the availability of multiple third-party storage services Concurrent upload: Upload to all available services at the same time, and whoever successfully uses it first, greatly improving the upload speed and success rate 5 Concurrent processing: Supports processing 5 videos simultaneously, which is 5 times faster than previous serial mode Zero configuration: No public network environment, no configuration, right out of the box Automatic fault tolerance: Automatic switching over for a single service failure to ensure the upload success rate 4. Built-in AI knowledge processing Not only extraction, but also processing. AI summary and rewriting: After filling in the API Key, by default, Ali Model (Tongyi Thousand Questions) is called to summarize, extract golden sentences, or rewrite subtitles. Model compatibility: Supports custom OpenAI format APIs and is compatible with mainstream models such as DeepSeek and ChatGPT. 5.📚Obsidian /Knowledge Base Friendly Designed specifically for knowledge management: Markdown export: Automatically generate subtitles into a neatly formatted Markdown file. YAML Frontmatter: Extracted video metadata (title, UP main, partition, label, etc.) is automatically written to the YAML area of the file header. Seamless integration: After downloading, you can directly drag it into Obsidian to use Bases management, or directly feed it as a corpus to LLM to build a personal knowledge base. ˇUsage suggestions In order to achieve a "silky smooth" experience, it is highly recommended to prepare the following two items before use: Alibaba Cloud DashScope API Key Core functions: Used for speech transcriptions (Paraformer) and AI summaries (Tongyi Thousand Questions). Linkage mechanism: If the large model API Key is not configured separately, the system will reuse this Key by default. If this Key is invalid or cleared, the large model function will be deactivated simultaneously (the status light turns red). It is recommended to pre-charge a small balance (such as 1-5 yuan). Station B Cookie (SESSDATA) Important: Station B has restrictions on API calls for users who are not logged in, and some self-contained subtitles need to be logged in to obtain correctly. Intelligent status feedback: The interface will display the validity of cookies in real time. If the Cookie fails, the system will automatically downgrade the policy (API Key needs to be configured) and change it to full voice transcriptions. Configure local direct chain services (optional but recommended) If you have a public IP or domain name (which supports HTTP and HTTPS), it is highly recommended to enable and configure it in the settings. This will bypass the instability of third-party direct-chain services and achieve extremely fast transcription. Interface prompt: The dynamic prompt bar below the input box will clearly inform you whether it is currently "local direct link (high speed)" or "third-party direct link (slow speed)", whether it is possible to "directly extract" or "voice transcribe", and your own expenses. Control your own expenses. Advanced gameplay 🎯Prompt Engineering Not just a simple summary! You can customize System Prompt in configuration to achieve unlimited possibilities: Golden sentence extraction: Please extract no less than 5 golden sentences or core ideas that appear in the video. Study Notes: Please summarize this content in Cornell's note format, including core keywords, summary and outline. Translator: Please translate the subtitles into English and summarize the key points. One-sentence summary: Please summarize the main theme of the video in one sentence, which is suitable for quick screening of videos. Batch productivity Collection extraction: Support directly pasting site collection/playlist links, and grab all videos in the entire series with one click. Guest account restrictions: Built-in Guest guest mode, all guests share 5 concurrent slots (queue up if they exceed them), making it easy to share the experience with friends. The visitor's history only exists in the browser session and disappears when refreshed. ˇ Precautions (Caveats)️ Comparison of operating modes Local direct link (recommended): Support HTTP and HTTPS, the system will turn on direct link mode, support 8 concurrent extraction, extremely fast, and connect directly to Alibaba Cloud. Third-party direct chain mode: Using built-in third-party direct chain service transfer, it now supports 5 concurrent processing and intelligent concurrent uploading. The speed is greatly improved and suitable for personal use, but it is still much slower than the local direct chain mode. Cookie validity period Station B cookies expire regularly (usually half a month to a month). When you encounter an extraction failure (especially a 412 error), please check the Cookie status indicator in the configuration as soon as possible, or re-copy the latest Cookie. data security and privacy Your API keys and cookies are encrypted and stored in the local database of the server you deploy (data/bilisub.db). It will never be uploaded to any third party and will not be collected by the project author. If you are using it on a public computer, it is recommended to use browser-free mode, log in to the guest account, and click "Clean Up Slow" at the bottom of the page after using it.