Skip to main content
Guide7 min read·Updated October 31, 2026
🧩

Best AI Agent Skills for Subtitles, Captions and Short Video (2026)

B

A. Frans

Published October 31, 2026

Agent SkillsSubtitlesVideoRemotionShort-Form

Captions used to be a finishing chore: export audio, wait for a transcript, fix the names, burn them in. Agents now drive that whole pipeline, and a growing number of open-source skills are built for it. Transcribe a recording, split the lines sensibly, translate them, and render a vertical clip, all from one conversation in your terminal.

I went through the video skills in our directory and kept the ones that touch subtitles, captions or short-form output. Some do the transcript, some do the editing, and some build the video from scratch in code. They're not interchangeable, so the comparison below shows which stage each one owns.

A note on expectations. Everything here is open source, and several are community projects with an "unreviewed" security status in our data. Video tools also handle large files and sometimes call external APIs, so the trust notes near the end matter more than usual.

Quick comparison

SkillWhat it doesStageBest forTrust
VideoCaptionerLLM-assisted subtitle generation, line splitting, correction, translationTranscript + subtitlesSubtitling a finished videoVerified
BaocutDrives the BaoCut macOS app CLI: transcribe, subtitle, translate, cutTranscript + rough cutMac users editing by commandCommunity
OpenChatCutConversational video editor with a multi-track timeline and Remotion renderingEditingEditing by chat, locallyVerified
Remotion VideoProgrammatic video with ReactRenderingData-driven and templated clipsVerified
Short Video MakerCreates TikTok, Reels and Shorts clips via MCP and a REST APIGenerationFaceless short-formVerified
Video TalkcraftVoiceover-driven explainers with word-level sync and motion recipesGenerationExplainers with kinetic textCommunity
OpenMontageAgentic video production with many pipelines and toolsFull productionEnd-to-end automationVerified
Native Subtitle Quote ImageGrabs frames keeping embedded subtitles, makes a 3:4 social imageStillsQuote cards from videoCommunity

VideoCaptioner: the subtitle specialist

VideoCaptioner is the most direct answer to "I have a video and I need good subtitles". Per its description it handles the whole subtitle flow with LLM help: generation, sentence splitting, correction and translation. Around 16,400 stars, verified tier, community-reviewed, and the license is GPL-3.0, which matters if you plan to embed it in something you distribute. Install with claude skill add WEIFENG2333/VideoCaptioner.

The project's title and summary are bilingual Chinese and English, so expect parts of the documentation in Chinese. That isn't a flaw, but check that the interface and prompts handle your target language before you rely on it for a client job. Line splitting and correction are the steps that separate usable captions from raw speech-to-text output, and they are where an LLM earns its place.

Baocut: subtitles from the command line, on a Mac

Baocut is an agent skill that drives the BaoCut macOS app's CLI, covering transcribe, subtitle, translate and cut. Our listing shows about 1,600 stars, MIT, community tier, unreviewed, and its star count climbed sharply in the last week of our data. Install with claude skill add JimLiu/baocut.

The catch is in the description: it needs the BaoCut app on macOS. If you're on Windows or Linux this one is out. If you're on a Mac and already edit with that app, the skill turns "cut the dead air and add Indonesian subtitles" into one request. Because it's a wrapper over a separate app, review what commands the skill is allowed to run before you grant it.

OpenChatCut: edit by chatting

OpenChatCut is an open-source, local-first conversational video editor. The listing describes a professional multi-track timeline, agent skills, MCP integration and Remotion rendering. About 2,800 stars, verified tier, community-reviewed, AGPL-3.0. Install with claude mcp add OpenChatCut -- npx -y 0xsline/OpenChatCut.

AGPL is the license to read twice. If you modify the software and offer it to others over a network, you owe them your source. For personal or internal use that's rarely an issue. For a SaaS product it can be. The local-first design is the useful part: your footage stays on your machine unless you send it somewhere yourself.

Remotion Video: when captions are code

Remotion Video generates programmatic video with React. Our listing: about 69,000 stars, MIT, pricing marked freemium, verified. Install with claude skill add remotion-dev/remotion.

For captions, Remotion's strength is control. You can render word-by-word highlighted subtitles, brand-matched fonts and animated text from a transcript file, then re-render in a different aspect ratio by changing a few lines. The "freemium" label deserves a click: check Remotion's own licensing terms before using it commercially in a company, because the framework's license has conditions for larger teams. If you want a head-to-head with two alternatives, see Remotion vs OpenMontage vs Short Video Maker.

Short Video Maker and Video Talkcraft: generating, not editing

These two start from text instead of footage.

Short Video Maker creates TikTok, Instagram Reels and YouTube Shorts clips through MCP and a REST API, with Remotion in its feature list. About 1,450 stars, MIT, verified. Install with claude mcp add short-video-maker -- npx -y gyoridavid/short-video-maker. One caution from our data: its last update is mid-2025, so check the repo for open issues and whether it works with current dependencies before you build on it.

Video Talkcraft targets voiceover-driven explainers. The listing mentions word-level voiceover sync, 78 motion recipe cards, a camera system meant to avoid slideshow-style videos, and Remotion rendering. About 2,000 stars, community tier, unreviewed, license field unasserted. Install with claude skill add Vincentwei1021/video-talkcraft. Word-level sync is what makes captions land on the right beat, so it's the one to try if your videos are narrated explainers rather than talking heads.

OpenMontage: the everything option

OpenMontage describes itself as an agentic video production system with 12 pipelines, 52 tools and more than 500 agent skills. About 73,000 stars, AGPL-3.0, verified, community-reviewed. Install with claude skill add calesthio/OpenMontage.

That breadth is both the pitch and the risk. A system with that many parts has a large surface to audit, and the self-description is the project's own marketing, not our verdict. Start by running a single pipeline on a throwaway clip, see what it calls out to, and only then widen scope. For a narrow job like captions alone, VideoCaptioner is a lighter starting point.

Native Subtitle Quote Image: a niche but handy one

Native Subtitle Quote Image takes frames from a video while keeping the subtitles already burned in, and produces a 3:4 social image. The listing's description is in Chinese; about 2,750 stars, MIT, community tier, unreviewed. Install with claude skill add chengyi-ai/native-subtitle-quote-image. Useful if you post quote cards from interviews and want the exact frame with its on-screen text. Make sure you have the rights to the footage you pull stills from.

Picking by job

  • Subtitle a finished video: VideoCaptioner. Mac user who edits in BaoCut: Baocut.
  • Edit by conversation, locally: OpenChatCut.
  • Branded, templated captions at scale: Remotion Video.
  • Faceless Shorts from a script: Short Video Maker, or Video Talkcraft for narrated explainers.
  • Hands-off pipeline across stages: OpenMontage, tested on one clip first.

If your work is repurposing long content into clips, our content repurposing roundup covers the surrounding workflow, and video editor skills covers editing tools more broadly. For the tool side rather than skills, see AI tools for video editors and content creators.

Trust notes

Three of the eight here are community tier and unreviewed: Baocut, Video Talkcraft and Native Subtitle Quote Image. Baocut also shells out to a separate app, which adds a second thing to trust. Video skills often run ffmpeg, download media, or call a hosted transcription or LLM API. Before installing, check three things in the repo: which commands the skill runs, whether it uploads your audio anywhere, and what the license requires. Three entries here are AGPL or GPL, so licensing can bite if you build a product on top.

Also mind consent. Auto-captioning someone's voice, or translating and republishing their video, needs their permission in most settings. Our data can't tell you about that, and neither can the skill.

Stars, update dates and tiers reflect our data in late October 2026 and will have changed.

FAQ

Which skill makes the most accurate captions? Accuracy depends on the speech model underneath, not the skill wrapper. VideoCaptioner's contribution is splitting and correcting lines with an LLM after transcription. Test on a minute of your own audio, with your own accents and terms.

Can I run all of this offline? OpenChatCut calls itself local-first, but transcription or translation steps in many of these can use hosted APIs. Check each repo's configuration for what stays on your machine.

Do these work on Windows? Baocut needs macOS. The others vary, and several depend on ffmpeg and Node or Python being installed. The repos list requirements.

Is Remotion free? Our listing marks it freemium. Read Remotion's current license terms for your company size before commercial use.

Are subtitle translations publishable as they are? Treat them as a draft. Names, jargon and idioms are where machine translation slips, so have a fluent speaker skim anything public.

Share this article

📬

Get More AI Tool Guides

New comparisons and guides every week. Join thousands of professionals staying ahead of the AI curve.