explaingit

louisedesadeleer/cut-video

Analysis updated 2026-06-24

20Audience · generalComplexity · 2/5Setup · moderate

TLDR

Claude Code skill that adds a /cut-video slash command to strip silences, ums, and dead air from long recordings on Apple Silicon while keeping laughs and comedic pauses intact.

Mindmap

mindmap
  root((cut-video))
    Inputs
      Long video file
      Tone preset
    Outputs
      Trimmed video
      Cut plan preview
    Use Cases
      Tighten podcast recordings
      Clean up interview footage
      Edit personal vlogs
    Tech Stack
      Claude Code
      ffmpeg
      Whisper
      VideoToolbox
    Presets
      playful
      sentimental
      documentary
Click or tap to explore — scroll the page freely

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

What do people build with it?

USE CASE 1

Tighten a 4K interview recording on a Mac without sending video to the cloud

USE CASE 2

Cut filler words from a podcast video while keeping laughs and comedic timing

USE CASE 3

Run a documentary preset that only removes silences longer than eight seconds

USE CASE 4

Preview the proposed cut plan before any frames are rendered

What is it built with?

Claude CodeffmpegWhisperVideoToolboxShell

How does it compare?

louisedesadeleer/cut-videoabolix/xplexabuzar-ansarii/androidlinux-gpu
Stars202020
LanguageGoShell
Setup difficultymoderatehardmoderate
Complexity2/53/52/5
Audiencegeneralops devopsdeveloper

Figures from each repo's GitHub metadata at analysis time.

How do you get it running?

Difficulty · moderate Time to first run · 30min

Built for Apple Silicon with VideoToolbox, Linux or Windows users need to remove the hardware acceleration flags.

In plain English

Cut Video is a skill that plugs into Claude Code, the coding assistant from Anthropic. Once installed, it adds a /cut-video slash command that takes a long recording and produces a tighter version with silences, ums, fillers, and dead air removed. The author tuned it to keep laughs and comedic pauses, so the cuts do not flatten joke timing or emotional beats. The skill works in four steps. First, it transcodes the source file into a 1080p H.264 proxy using hardware decoding, so each later step runs in seconds instead of minutes. Then it runs OpenAI's Whisper model (the tiny English variant) to transcribe the audio with word-level timestamps. Next it builds a cut plan using a tone-aware silence heuristic, drops fillers and false starts, and shows you that plan before changing anything. Finally it renders the result with ffmpeg using a trim and concat approach, which the README says keeps audio in sync where the more common select filter tends to drift. The README explains why the project exists. Most silence-cutters either cut too aggressively (removing laughs) or are slow on large 4K HEVC files. The author tested against a 7-minute 4K interview that took over 5 minutes just to decode in other tools, and brought the same cut down to about 30 seconds by combining VideoToolbox hardware acceleration with hardware-friendly trim and concat steps. Three tone presets are offered: playful (the default, aggressive on dead air but keeps comedic beats), sentimental (gentler, preserves longer pauses for emotional moments), and documentary (only cuts genuine dead air over 8 seconds). The skill runs entirely on your machine with no cloud calls, and is built for Apple Silicon on macOS. Linux or Windows users would need to remove the VideoToolbox flags. The README is clear that this is a cutting tool only. It does not add zooms, captions, b-roll, or motion graphics. It needs Claude Code, ffmpeg with libx264, and Whisper installed first.

Copy-paste prompts

Prompt 1
Install cut-video as a Claude Code skill and run it on my latest interview recording with the playful preset
Prompt 2
Adapt cut-video to run on Linux by replacing the VideoToolbox flags with software decoding
Prompt 3
Tune cut-video's silence threshold so a 7 minute talk keeps every pause longer than two seconds
Prompt 4
Add captions to the output of cut-video by chaining a second skill that uses the Whisper transcript
Prompt 5
Explain why cut-video uses trim and concat instead of the ffmpeg select filter for audio sync

Frequently asked questions

What is cut-video?

Claude Code skill that adds a /cut-video slash command to strip silences, ums, and dead air from long recordings on Apple Silicon while keeping laughs and comedic pauses intact.

How hard is cut-video to set up?

Setup difficulty is rated moderate, with roughly 30min to a first successful run.

Who is cut-video for?

Mainly general.

Open on GitHub → Explain another repo

This repo across BitVibe Labs

Verify against the repo before relying on details.