Clips
Six-step video pipeline: download, transcribe, cut, translate, karaoke HTML
Test report
- Verdict
- Works with setup
- Score
- Tested
- Aug 7, 2026
- Environment
- Claude Code 2.x (agent harness)
Ran the cut and translate steps on a 120-second fixture with a 22-sentence transcript and measured both against a no-skill pass at the same task. The skill's insistence on re-encoding produced a clip with zero negative-timestamp packets, while the obvious -c copy shortcut I reached for without it carried 233 of them, roughly nine seconds of pre-roll a browser player has to discard through the edit list. Subtitles were the sharper gap: writing the file the natural way produced seven bare strings in the cn array and a clip key instead of clip_num, exactly the shape the step manual warns the karaoke renderer rejects with 'invalid type: string, expected struct', whereas following the documented schema and its character-count interpolation gave 13 well-formed cues with no timing gaps. The problem is that none of that guidance is inside the installed skill: the copied folder is a 48-line SKILL.md that shells out to nf-guide for every step, a binary you only get by cargo-building NextFrame's seventeen-crate Rust workspace, on top of yt-dlp, WhisperX and the nf CLI for the final HTML. End to end was never verified here, the SKILL.md is Chinese-only with no install or dependency notes, and its own step list skips from 04 to 06.
Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works
- Installs cleanly 3/5
- Triggers reliably 5/5
- Output vs. baseline 7/10
- Docs & honesty 2/5
What Clips does
Drives a six-step short-form video pipeline that downloads a source video, transcribes it with WhisperX, picks highlight segments, cuts them with ffmpeg, translates the subtitles and renders a word-synced karaoke HTML player. It triggers when you ask to cut highlight clips from a long video or podcast, or to produce subtitled shorts with Chinese captions. Each step's operating manual is fetched from the nf-guide binary that ships in the NextFrame Rust workspace, so the skill folder on its own is only the state machine.
How to install Clips
git clone https://github.com/ChaosRealmsAI/NextFrame.git
mkdir -p ~/.claude/skills
cd NextFrame && cp -r crates/nf-agent/skills/clips ~/.claude/skills/clips
Skills live in ~/.claude/skills/ (global) or .claude/skills/
(per-project). Restart Claude Code after installing.
Commands — how to trigger Clips
-
/clipsSix-step video pipeline: download, transcribe, cut, translate, karaoke HTML
It also activates on plain-language prompts like these:
-
Cut highlight clips from this podcast and add Chinese karaoke subtitles -
Download this video, transcribe it, and cut the best 30-second clip -
Turn this long video into subtitled shorts with translated captions
Frequently asked questions
- Is the Clips skill free?
- Yes. The skill itself is free from ChaosRealmsAI/NextFrame. SkillProof publishes the install command and an independent test verdict at no cost.
- Does Clips work with Claude Code?
- We tested it with Claude Code 2.x (agent harness) on Aug 7, 2026. Verdict: Works with setup. Ran the cut and translate steps on a 120-second fixture with a 22-sentence transcript and measured both against a no-skill pass at the same task. The skill's insistence on re-encoding produced a clip with zero negative-timestamp packets, while the obvious -c copy shortcut I reached for without it carried 233 of them, roughly nine seconds of pre-roll a browser player has to discard through the edit list. Subtitles were the sharper gap: writing the file the natural way produced seven bare strings in the cn array and a clip key instead of clip_num, exactly the shape the step manual warns the karaoke renderer rejects with 'invalid type: string, expected struct', whereas following the documented schema and its character-count interpolation gave 13 well-formed cues with no timing gaps. The problem is that none of that guidance is inside the installed skill: the copied folder is a 48-line SKILL.md that shells out to nf-guide for every step, a binary you only get by cargo-building NextFrame's seventeen-crate Rust workspace, on top of yt-dlp, WhisperX and the nf CLI for the final HTML. End to end was never verified here, the SKILL.md is Chinese-only with no install or dependency notes, and its own step list skips from 04 to 06.
- What is the Clips SkillProof Score?
- 6.8/10 — installs cleanly 3/5, triggers reliably 5/5, output vs. baseline 7/10, docs & honesty 2/5.
- How do I install Clips?
- Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.
- Can I use Clips with Cursor, Copilot, Gemini CLI, Codex or other AI tools?
- The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.