Kevin Stratvert sets up a ChatGPT dot as a YouTube coach: it connects vidIQ, sets a 20% daily-views goal, shares video ideas in a Space and keeps checking in on its own.
Turn a raw recording into a polished video short with a free AI skill
Peter Yang's free video-short skill takes a raw talking-to-camera clip and returns a short with B-roll, motion graphics, captions and AI-generated video, refined through frame-by-frame feedback on a review page.
Peter Yang · YouTube · published Oct 21, 2026 · Content & Creative
▶ Watch on YouTubePrompt
/video-short Here is my raw recording: [file name, e.g. raw.mp4] in my Downloads folder. Edit it into a short that is as visually impressive as possible, with captions, motion graphics, B-roll you find online, and AI-generated video and music. First give me a scene-by-scene plan (speaker full screen, B-roll, split view, graphic) and the B-roll sources you found, so I can give feedback before you render.
Visual breakdown
Peter Yang's free video-short skill turns any raw talking-to-camera recording into a polished short with B-roll, motion graphics, captions and AI-generated video. The AI drafts a scene-by-scene edit, you leave feedback on a review page, and it re-edits until you approve the render.
The problem: Short video is a core marketing skill on every platform, but editing a raw recording into a polished short takes a lot of manual work.
From raw recording to polished short
- Install the skillPaste the GitHub URL into the agent; it checks for ffmpeg, yt-dlp and hyperframes and offers to install them
- Raw recordingA 54-second clip of him talking to the camera
- Scene-by-scene planAI decides where to cut and lays out scenes: speaker full screen, B-roll, graphic split viewDropped: Pauses, retakes and filler words (basic cutting built in)
- First-cut editAdds captions, HTML motion graphics, online B-roll, and Gemini Veo video and Lyria music
- Review pagePlay the short and leave comments scene by scene, then send notesDropped: Fixed on the next round, e.g. a cut-off head
- Approve and renderFinal clip ready to upload to Instagram, YouTube and more
Setup / inputsYou reviewAI agentResult
Four scene types the skill cuts between
- A-roll: him talking to the camera
- Punch: a tighter shot
- Split: speaker on the bottom half, B-roll on top
- Full-screen B-roll
By the numbers
4steps in the workflow
54 secraw recording
200BShorts views a day, used as an overlay
Do it yourself
He uses Antigravity with Gemini 4 Argon, but says any frontier model works with the skill.
- Paste: install the video-short skill globally from github.com/petergyang/video-short
- Let it check for ffmpeg, yt-dlp and hyperframes and install what's missing
- Optional: create a Gemini API key in Google AI Studio and turn on billing; ask the agent to open the .env file so you paste the key there, not in chat
- Run /video-short with the path to your raw recording
- Review the scene plan and B-roll sources, and give feedback
- On the review page comment per scene, send notes, repeat, then approve and render
It isn't one-shot: his clip took a few more rounds of feedback to reach the version shown.
Tips and caveats from the video
- Never paste an API key into the chat; have the agent open the .env file and paste it there.
- Everything is free except the optional Gemini key; AI video and music cost a cent or a few cents, so review the paid steps first.
- The skill file carries short-form best practices: hook in the first 3 seconds, visuals that show what the words don't, real footage first, nothing too tiny for a phone.
- He built an eval so the AI checks its own work, such as whether the hook lands in 3 seconds, before showing you.
- For longer videos, a prompt that shortens pauses to 1 second or less and strikes retakes and filler words is a lifesaver.
Chapters
0:00The free video short skill0:50Four steps overview1:29Step 1: install the skill2:45Setting up the Gemini API key4:12Step 2: upload and get a draft edit5:51Prompt to cut pauses and filler words7:43Step 3: frame-by-frame feedback10:16Approve and render final10:54What the skill can do11:43Best practices in the skill file
Distilled from the video’s captions
How to try it
- Open Gemini.
- Copy the prompt and fill in the [placeholders].
- Watch Peter Yang's video for the full walkthrough and setup.
Related use cases
Futurepedia shows how Claude writes plain JavaScript that draws every frame, sound effects included, then uses his free 86-style atlas to make an explainer, even synced to a voiceover.
Mikey builds Lumen, a liquid-glass journal app, in six Gemini 4 Argon prompts on Base44: tags and search, AI reflection, sign-up, Stripe premium and a final polish.
Tim builds a Python AI agent from scratch: you speak what you want, and a 6-step pipeline combining text, voice, image and video generation via OpenRouter produces a fully scripted video.