Self-hosted AI distillation · static, instant

Turn fragmented videos into structured knowledge you actually keep

No more information anxiety or bookmark graveyards. On-premise speech recognition (SenseVoice/Whisper) plus LLM distillation extract core insights, Mermaid mind maps and an actionable checklist — automatically.

⚡ Millisecond offline WASM search 🎙️ Millisecond timestamp seeking 🗺️ Native Mermaid mind maps
deep-read notes
562
deep-read notes
knowledge categories
6
knowledge categories
bilingual
中⇄EN
bilingual

From one video to searchable knowledge

The whole pipeline is self-hosted and re-runnable — every step is logged and traceable.

See the full pipeline →
  1. 01

    Speech recognition

    SenseVoice / Whisper transcribes on the NAS, sentence by sentence with timestamps.

  2. 02

    LLM distillation

    The model strips filler, splits chapters and extracts insights, principles and actions.

  3. 03

    Structured output

    Mermaid mind map plus checklist, with the English edition and narration derived alongside.

  4. 04

    Offline full-text search

    A WASM index runs in the browser — no backend call, instant results.

RSS