skills.rest
TeamWiseFlow
TeamWiseFlowOFFICIAL
Explore all skills by this author →
GitHub

talking-head-cut

Detects filler words, silences, and highlights in talking-head videos via ASR timestamps to generate cut plans.

8.5k|1.4k|Updated Apr 24, 2024
Search Tags:#FFmpeg#video editing#asr#filler removal#talking head#highlight detection#cut plan
One-click install
npx skills add https://github.com/TeamWiseFlow/wiseflow --skill talking-head-cut
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: talking-head-cut
Source: https://github.com/TeamWiseFlow/wiseflow/tree/main/crews/main/skills/talking-head-cut
Command: npx skills add https://github.com/TeamWiseFlow/wiseflow --skill talking-head-cut

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires requests, and includes scripts (resource) components.

What problem does it solve?

Editing talking-head videos (vlogs, speeches, interviews, livestreams) manually is slow: you must scrub hours of footage to remove filler words, stutters, and dead air, or to find the highlight moments worth clipping. This Skill automates the detection phase by transcribing speech with word-level timestamps and producing a structured cut plan.

Core Features & Use Cases

  • Filler Removal: Detects filler words (嗯/呃/um/uh), silences, stutters, false starts, and repeated sentences, marking them for removal in a cut_plan.json file.
  • Highlight Extraction: Identifies high-density, novel speech segments and marks them as keep segments for short highlight reels.
  • Human-in-the-loop Gate: The cut plan must be reviewed and confirmed by the user before the actual ffmpeg cutting is executed via the video-edit skill.
  • Use Case: You have a 30-minute interview recording and want a clean 60-second highlight clip. Run detection in highlight mode, review the generated cut_plan.json, then apply the cut and verify the result with the video-review gate.

Quick Start

Ask the agent to remove filler words and silences from your talking-head video, or to cut it into a highlight reel of a target duration.

Frequently Asked Questions about talking-head-cut

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I remove filler words from a video automatically?▼

Run the detection script with mode filler to transcribe the audio via ASR and generate a cut_plan.json marking filler words, silences, stutters, and repeats for removal. After confirming the plan, the video-edit apply-cut command performs the actual ffmpeg cutting.

How to create a highlight reel from a long talking-head video?▼

Use highlight mode, which keeps only segments with high speech density and high novelty relative to the full transcript. The resulting cut plan lists the highlight segments, which are then extracted and concatenated with short fades into a single clip.

What ASR service does the transcription use?▼

Transcription uses Volcano Engine's Doubao speech recognition API, configured through VOLC_ASR_APP_ID plus VOLC_ASR_ACCESS_KEY or a single VOLC_ASR_APP_KEY environment variable. Without these credentials the script exits with code 2.

Does it support English videos or only Chinese?▼

Both languages are supported through the language parameter. Chinese uses fillers like 嗯, 呃, and 额, while English uses um, uh, uhm, and er, with density thresholds adjusted per language.

When should I not use this skill for video editing?▼

Do not use it when the video has no speech or when highlights depend on visuals rather than spoken content; use a visual editing workflow instead. It also does not burn subtitles, add background music, or generate videos from scratch.

Related Skills

Explore More Content & Communication Skills→
whyiloveher2411whyiloveher2411

talking-head-recut

Overlay transcript-synchronized graphic cards onto talking-head videos.

Community
Intermediate
dontbesilent2025dontbesilent2025
9.8k

dbs-content-risk-check

Reviews social media content for platform moderation triggers and substantive violations before publishing.

Community
Advanced
petbrainspetbrains
13

frontend-debug-linting

Lint, type-check, format, and verify React/Next.js projects in browsers.

Official
Intermediate
TheArchitectitTheArchitectit
3

video-frames

Extract and analyze video frames using ffmpeg, ffprobe, and Pillow.

Community
Intermediate