User repository on GitHub
按个人偏好与目标时长剪辑长视频高光：Jev 评分、可追溯选片、本地 FFmpeg 粗剪。AI agent skill with evidence-based highlight selection.
- Forks
- 4
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Agents & MCP
Jev Highlight Cutter
An agent skill plus Python CLI that turns long interviews and podcasts into evidence-based highlight reels: the agent reads the transcript, Jev scores the text evidence, and local FFmpeg cuts a rough cut that fits your taste and target duration.
- Stars
- 15
- Language
- Python
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- gnipbao
- Last verified
- Oct 2026
Highlights
- A Codex-style agent skill with a Python CLI: the agent reads the transcript and reviews cut boundaries, Jev scores the text evidence, and local FFmpeg cuts.
- Reads SRT, VTT and Whisper-style JSON while keeping original absolute timestamps, then picks non-overlapping segments inside a target-duration budget.
- Keeps scores, selection rationale and source-to-cut time mapping, and renders a clickable local review page, MP4 clips and an export verification report.
- Requires Python 3.10+ and FFmpeg with ffprobe; the core CLI uses only the Python standard library, so no pip install is needed.
- MIT licensed; real scoring calls a user-global TYPESAFE_API_KEY, while doctor runs without calling Jev or verifying account quota.
Quickstart
bash
mkdir -p "$HOME/.codex/skills"
git clone https://github.com/gnipbao/jev-highlight-cutter.git "$HOME/.codex/skills/jev-highlight-cutter"
python3 "$HOME/.codex/skills/jev-highlight-cutter/scripts/highlight.py" doctorWatch out
README reviewed 2026-10-17; the digest truncates inside the first manual CLI example, so the command sequence after the run-directory setup was not captured. The author scopes it to interview podcasts and states the output is a rough cut: no bilingual subtitles, dubbing, BGM or JianYing/CutChat project export, and no benchmarks are published.
Questions about this entry
Short answers drawn from the linked project or page.
An agent skill plus Python CLI from gnipbao that turns long interviews, podcasts and livestreams into evidence-based highlight reels; the agent handles content understanding and cut-boundary review while Jev scores transcript evidence and local FFmpeg does the cutting.
Jev scores the text evidence: transcripts in SRT, VTT or Whisper-style JSON keep their absolute timestamps, candidate segments are evaluated for information value, opening appeal, completeness and taste match, and non-overlapping picks inside your target duration survive.
Create ~/.codex/skills, clone the repository into a jev-highlight-cutter folder there, and run scripts/highlight.py doctor to check the environment; other agents that read SKILL.md can point at the repo from their own skill directory.
Python 3.10+, Git and FFmpeg including ffprobe; the core CLI depends only on the Python standard library, so there is no pip step. macOS Homebrew and Ubuntu apt commands are documented, and Windows runs the Bash syntax through WSL.
Real scoring uses a user-global TYPESAFE_API_KEY read from the environment, not a repository .env file. Doctor does not call Jev or check account quota; without a key you can still build the index, prepare candidates and inspect the pending request.
No, it outputs a rough cut: a clickable local review page, standalone MP4 clips, an original-audio compilation and an export verification report. Bilingual subtitles, dubbing, BGM and JianYing/CutChat project export are left to later fine-cutting tools.
The digest truncates inside the first manual CLI example, right after the run-directory setup, so the remaining commands, API configuration details, privacy and cost sections and the validation limits were not captured; check docs/jev-api.md, references/commands.md and references/validation.md before relying on it.
More Jev resources like this
An independent agent skill and Python CLI for local video remixing: shot-aware slots for beats and original narration are filled under Jev judgments and constrained orchestration, then rendered into a reviewable cut with originality and asset-license reports.
Jev Agents & MCP#community#github#python
Communitygnipbao/jev-remix-cutterPython
gnipbao/jev-remix-cutterPythonA Claude skill and API from tomascupr that turns any video link into one typed JSON document: summary, chapters, scenes, cast, key moments, the emotional arc, brands, on-screen text, and the full transcript. Jev then judges those fields so your code can branch on probabilities.
Jev Agents & MCP#community#github#claude
Community
tomascupr/reelqlHTML
tomascupr/reelqlHTMLA zero-embedding approach to video intelligence from DeepAleph: jevreel lays a video out on labelled contact sheets, one frame per second and sixty frames to an image, and has Jev judge every sheet in parallel. Your coding agent gets moments, times and probabilities back over MCP.
Jev Agents & MCP#community#github#mcp
Community
deepaleph/jevreelPython
deepaleph/jevreelPythonLocal-first MCP memory server for IDEs: tree-sitter indexing into a local ChromaDB store, plus an optional Jev layer that screens retrieved passages for relevance, evidence, contradiction and prompt injection.
Jev Agents & MCP#community#github#python
Community
KikeVen/zerikai_memoryPython
KikeVen/zerikai_memoryPythonAn auto-clipper that turns long videos into short standalone clips: code enumerates every plausible cut point, Jev judges the clip each candidate edge would make, and you get ranked mp4s for Shorts, Reels and TikTok.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
vbs2004/jevcutPython
vbs2004/jevcutPythonA local System One decision layer for AI coding agents, shipped as an MCP server and Agent Skill, that offloads high-volume classification, checks, scores and ranking from frontier LLMs to sub-50ms local models with calibrated output.
Jev Agents & MCP#community#github#mcp
Community
aktonay/snapdecPython
aktonay/snapdecPythonFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.