How to use the KrillinAI Vertical Video Renderer agent skill

What an agent skill is and what this one makes the agent do

An agent skill is a folder with a SKILL.md file that gives an AI agent specific instructions for one kind of task. When the task matches the skill's description, the agent reads those instructions and follows them.

Once this skill is installed, the agent knows how to run the render-vertical command of the KrillinAI CLI. It can convert a source video to portrait format, add short bilingual subtitles and titles, render a dubbed vertical version, and run basic checks. The skill does not do the video work itself; it drives the KrillinAI tool.

• Convert a source video to vertical format (transferred_vertical_video.mp4)
• Add short bilingual subtitles and a major and minor title
• Render a dubbed vertical video with the --dubbed option
• Check the output: extract a preview frame and confirm the JSON reports success

Install the krillinai-render-vertical skill

Download the skill ZIP from the page where you found this guide, then unzip it. The skill folder should contain SKILL.md. Place the folder in one of these locations, depending on how you want to use it:

For a personal setup available across projects, put it in ~/.claude/skills/. For a single project, put it in that project's .claude/skills/ folder.

If you use the Claude apps instead, open Settings, go to Capabilities, then Skills, and upload the skill there.

• Personal: ~/.claude/skills/
• Project only: .claude/skills/ inside your project folder
• Claude apps: Settings → Capabilities → Skills → upload

Set up the requirements before you run anything

The skill depends on several tools. It needs the KrillinAI CLI binary, whose location is defined in skills/krillinai-cli/references/cli-contract.md. That reference file is not included with this skill, so you need it from your own KrillinAI setup. The contract file also covers the execution working directory, configuration, and JSON behavior.

You also need ffmpeg, which is used to extract preview frames, and rg (ripgrep) for text checks. Your input video and SRT subtitle files must be in a working directory.

• KrillinAI CLI binary (location in cli-contract.md, not included)
• ffmpeg for preview frame extraction
• rg (ripgrep) for text checks
• Input video and SRT subtitle files in a working directory

Ask the agent for a vertical render in plain language

The skill triggers when your request involves rendering portrait video with KrillinAI, such as converting a video to vertical, adding short bilingual subtitles, or making a dubbed vertical version. You do not need to name commands. Be specific about your files and the titles you want.

Good requests name the source video, the subtitle file, and what you want the output to look like. For example: 'Turn origin_video.mp4 into a vertical video with the bilingual subtitle file and the titles Today Topic and AI Video.' The agent should then check the CLI setup, run the render, and report the result.

• Name the video and subtitle files you want used
• Say whether you want bilingual subtitles or a dubbed version
• Give the major and minor titles if you want them on screen

Run the bilingual and dubbed renders

The bilingual render takes a source video and a mixed-language SRT file, plus two title strings. The default final output is vertical_bilingual.mp4. The dubbed render uses a video with text-to-speech audio and a target-language SRT, and adds the --dubbed flag. Its final output is vertical_dubbed.mp4.

The skill's subtitle handling aims to keep the screen to one English line plus one Chinese line when possible. Chinese text is word-segmented before line decisions, and long Chinese subtitles are split across time rather than stacked.

• Default output: vertical_bilingual.mp4
• Dubbed output (with --dubbed): vertical_dubbed.mp4
• Intermediate file: transferred_vertical_video.mp4
• Final variants are reported in the manifest under the key vertical_video

Check the result before you publish

The skill asks for both text and visual checks. For text, search the formatted ASS file for literal \N line breaks, which the skill avoids for long Chinese subtitles. For visual, extract a frame from the rendered video to inspect subtitle readability.

Beyond those checks, the render should return terminal JSON with "ok": true, valid video and audio streams, readable subtitles, and titles that are not garbled. Watch the preview frame yourself before posting.

• Terminal JSON contains "ok": true
• Video and audio streams are valid
• Subtitles are readable and titles are not garbled
• Preview frame looks correct

Safety notes and tips

This skill runs scripts, so read SKILL.md and any files it references before installing it, as you would with any code you did not write. It does not use the network, according to its listing. It works on the files in your working directory and calls the KrillinAI CLI and ffmpeg.

For cleaner output, prepare your SRT files before rendering, keep the titles short, and always review the preview frame. The skill is most useful once your KrillinAI setup already works.

• Review SKILL.md and referenced files before installing
• The skill runs scripts but uses no network access, per its listing
• Keep titles short and check the preview frame every time
• Test a render on a short clip before a long batch