Web Video Auto Learner

Problem

Web video courses often split playback, pop-up answering, progress saving, audio capture, and summarization into manual steps. Lazy-loaded players, interrupting pop-ups, lost progress after browser crashes, and missing course transcripts make repeated learning fragile. web-video-auto-learner targets this workflow and uses Playwright, Tkinter, and AI models to turn web video study into a repeatable pipeline.

How it works

Core capabilities include:

• Automatic login: supports account/password login or Storage State, with login state bound to the local machine.
• Automatic playback: opens videos in play_list order and handles lazy loading in players such as xgplayer.
• Popup handling: scans MyLayer or quiz pop-ups, then closes them or calls AI-assisted answering when configured.
• Progress guard: saves playback progress periodically and attempts recovery after browser failures.
• Audio capture and transcription: records page audio via Web Audio API, then transcribes with iFlytek LFASR or local Whisper.
• AI summarization: calls SiliconFlow, DeepSeek, OpenAI, or Ollama to generate course notes; some reasoning models can output chain-of-thought.

The key steps are configuring the login page, playlist, AI provider, and transcription engine, then running console, simple, or playlist mode to complete playback, recording, transcription, and summarization.

Boundaries and notes

It is better suited to stable, accessible online course sites. AI features require valid API keys or a local model; audio capture requires the video to be playing; cloud transcription depends on iFlytek credentials, and local transcription depends on PyTorch, Transformers, and ffmpeg. Storage State is device-bound and should not be casually shared across machines.