Pick a motion from 168 live examples, and AI builds your video to match

Tetumemo packaged Motion Prompt as a skill, loaded it into Claude Sonnet 5.5, gave it their own explainer post on Grok Bot, and asked for a 30-second video. The result runs 33 seconds at 1920×1080. It's a promo for their own post, pointing readers to the prompt collection inside it. Tetumemo didn't say which motions from the dictionary were used. X

A first-month price of 「2,640円」 appears on a black screen as a blue circle slides in. Next come three cards for specialized bots, each with a narrow role. Arrows draw in to show work passing from one bot to the next, then lines connect the three bots to “you,” placed in the center like a conductor. The phrase 「たった1回 約3分」 (just once, about 3 minutes) appears with a timeline running from a request at 17:56 to a finished article draft at 18:00. It ends on 「貼るだけで使えるプロンプト 97個」 (97 prompts you can just paste in). The subtitles at the bottom change with each scene.

Tetumemo wrote: 「Claudeのモーショングラフィックス性能は非常に高いので、オススメです！」 (Claude is really good at motion graphics, so I recommend it!) X

Suzuki installed motion-prompt on October 17 and set up Claude Code and Codex to read the same dictionary. On October 21, Suzuki used Codex to make a new 12-second video (1920×1080, 30fps, no audio) for a note post. note

For the first 3.4 seconds, the two lines of 「かっこよく、を具体的に。」 (Make “cool” specific.) rise out of a mask from below, 0.6 seconds per line (#005 mask reveal). The middle section brings in navy and orange shapes, and starting at 4.55 seconds a rounded rectangle morphs into a circle over 0.58 seconds (#012 shape morph). The last four-plus seconds hold on 「名前が、依頼を変える。」 (A name changes the request.) until the end. Suzuki pulled frames at 1.40, 4.80, 5.60, and 9.50 seconds from the finished video for the post. note

Suzuki chose the shape to morph, the navy and orange palette, the length, the layout, and the two lines of copy. The roughly 240 points that form the shape's outline, the pulse that grows and shrinks by 7%, and the easing that overshoots slightly and settles back (backOut 1.6) all came straight from the values listed in #012, Suzuki wrote. The post ends with a link to AI consulting from Suzuki's company. note

Nagi asked Claude Sonnet, with no tools, for a title animation in a single HTML file. The scene shows the headline 「今日の特集」 (Today's feature) and the line below it, 「名前を知れば頼める」 (Know the name and you can ask for it), appearing. Nagi wrote the request three ways and ran each one twice. note

With only 「見出しがかっこよく出てくるようにしてください。」 (Please make the headline come in looking cool.), the first run had the letters shatter and settle alongside a glowing line, and the second had particles gather into the text. When the request named the motion (mask reveal) and gave 0.6 seconds, as below, both runs revealed the text from below over 600ms, with the second line following 150–160ms later. Pasting example #005's copy-ready prompt as-is reproduced the example's timing and even its coral and yellow colors. note

Nagi notes this was a small test: one model, Claude Sonnet, two runs per request.