Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

Boson-Video

Video summarizer that ties every line to its second; Jev checks each summary sentence against what was said, with numbers checked in code.

ProjectbetaMIT
View on GitHub ↗

Overview

Boson-Video helps you get the point of a long video without watching it all. Paste a YouTube link or drop a file and every line of the summary, transcript and on-screen text carries the second it came from. A ribbon shows the whole video with scenes, chapters and most-replayed parts; summaries come with frames; transcripts include English translations and explained terms; and questions are answered with cited moments, marking anything the video does not say as background. It targets long talks, lectures and finance videos in Chinese and English.

Jev is the checker: each summary sentence is checked against what was actually said and marked as confirmed, unsure or contradicted, while numbers are compared in code rather than by the model. It also runs as an MCP plugin for Claude, Cursor and Codex with nothing else to install, and the core runs locally. The README notes YouTube's terms restrict automated access for products.

What the decision model decides

TypeQuestionAnswers
Yes / noDoes the transcript support this summary sentence?(inferred)confirmed · unsure · contradicted

How it works

  1. Ingest a video and build a timed transcript and frames
  2. Generate a sectioned summary
  3. Jev checks each sentence against what was said; code compares numbers
  4. Mark sentences with check, question mark or cross
  5. Answer questions with cited moments

Strengths

  • Every claim tied to a timestamp
  • Numbers checked in code, not by the model
  • Usable as an MCP plugin

Caveats

  • YouTube terms restrict automated access
  • Checker needs a TypeSafe key

Getting started

claude mcp add boson-video -- uvx boson-video mcp

At a glance

Type
Project
Status
beta
Language
Python
Runtime
local app + MCP server
Has code
yes
License
MIT

Use cases and integrations

How curated lists describe it

  • Summarises a video with every line tied to the second it came from; Jev checks each summary sentence against what was said, with numbers compared in code.

    MrJev/awesome-jev

Frequently asked questions

What does Jev check?

Whether each summary sentence is supported by what was said in the video.

Who checks numbers?

Code, not the model.

Can I use it from Claude?

Yes, as an MCP server via uvx boson-video mcp.

Related projects

  • jevmory

    Local memory for coding agents that stores verbatim quotes and has Jev grade memory lines as keep, stale, wrong or unsupported.

    ProjectbetaPython
  • typesafe-as-a-judge

    Unofficial Codex and Claude Code MCP plugin giving agents bounded Jev tools for routing, ranking, extraction, verification and judging.

    MCP serverbeta
  • jev-got

    Game of Thrones roleplay where an LLM writes each scene and Jev labels location, beat, mood, danger and whether prose stayed in fiction.

    AppprototypeTypeScript
  • codex-gemini-jev-bridge

    Local MCP bridge letting Codex hand drafting to Gemini via a Google AI Pro quota, with Jev checking drafts against explicit criteria.

    MCP serverbeta
  • jev_project_context

    Evidence-first experiment-memory skill for coding agents, with an optional Jev audit layer that checks provenance and claim support.

    Agent skillbetaPython
  • jev-toolkit

    Small stdio MCP server with ask, rank and verify tools backed by Jev, a local event log and installers for several agent harnesses.

    MCP serverbetaTypeScript