Demo · Claude

Make a motion graphics explainer video from your script

AfterReady to post · 0:40
BeforeAs it came in · 2:11

A creator asks Claude to turn a 2-minute explainer that is one static slide into real motion graphics. cuttingfloor tightens the script to 40 seconds, reads it in a British voice, boards nine beats on the words and animates each one with camera moves, timed builds and sound effects.

LiveWatch it live now Kazi gives Big Buck Bunny a happier ending.Kazi is one way in: ChatGPT, Claude or any MCP app connects the same way.Free · a few minutes · no sign-up
Try it live in Kazi, free

Kazi is our own AI agent, on your phone or computer. It opens with a sample clip and edits it on cuttingfloor.

Your turn

Do this with your own video

Add cuttingfloor to the app you already use, then ask for the edit in your own words.

https://api.cuttingfloor.ai/mcp
  1. ChatGPT

    1. Open chatgpt.com/plugins, click +, then Add custom MCP server.
    2. Name it cuttingfloor, paste the address above and set Authentication to OAuth.
    3. Click I understand and want to continue, then Create as a plugin. Sign in when asked.
    4. In a chat, type @cuttingfloor and ask for the edit.
    5. Attach your video in the chat. ChatGPT hands it to cuttingfloor.
    Open ChatGPT plugins ↗
  2. Claude

    1. Open Customize → Connectors, click + Add, then Add custom connector.
    2. Name it cuttingfloor, paste the address above and click Continue.
    3. Click Add, then sign in on the page that opens.
    4. In a chat, open + → Connectors, turn cuttingfloor on and ask for the edit.
    5. Claude gives you a link to send your video from your phone or computer.
    Open Claude ↗
  3. Kazi

    1. Ask Kazi: “Connect the cuttingfloor MCP server at https://api.cuttingfloor.ai/mcp”. It works on your phone too.
    2. Ask for the edit. Kazi gives you a link to send your video from your phone or computer.
    Open Kazi ↗
01

Before

The clip as it came in · 2:11

memory-explainer.mp4 · 2:11

What needs to go

Most people give their assistant a notepad tool. The model stops mid-turn, emits a tool call, waits, then keeps going. We don’t give the assistant a tool. We give it a convention: it writes its notes inline, inside the response it’s already generating. The notes ride the response stream. No tool call, no round trip. A small state machine flips to recording on the open sentinel and done on the close. That’s working memory. Persistence happens later, at eviction. …three details that matter. Start there.

Struck in red: 2 things the agent cuts.

02

The edit

In Claude, with cuttingfloor connected

Most people give their assistant a notepad tool. The model stops mid-turn, emits a tool call, waits, then keeps going. We don’t give the assistant a tool. We give it a convention: it writes its notes inline, inside the response it’s already generating. The notes ride the response stream. No tool call, no round trip. A small state machine flips to recording on the open sentinel and done on the close. That’s working memory. Persistence happens later, at eviction. …three details that matter. Start there.

tip-1-motion-graphics.mp4 · 0:40
Claudecuttingfloor
  1. This explainer is one slide for two minutes. Turn tip one into real motion graphics: under a minute, British voice.

    I’ll tighten the script to about 40 seconds, record it in a British voice, then board it beat by beat on the words before animating anything.

    • Used cuttingfloor › skills.get storyboard before animate
    • Used cuttingfloor › skills.get make motion graphics
    • Used cuttingfloor › media.get read the old script
    • Used cuttingfloor › create.voiceover British voice, 38 s, every word timed
    • Used cuttingfloor › media.import Inter and JetBrains Mono
    • Used cuttingfloor › create.motion nine beats, one scene each
    • Used cuttingfloor › render.frames checked every beat
    • Used cuttingfloor › media.import a style frame of each beat
    • Used cuttingfloor › edit.batch the board, timed to the voice
    • Used cuttingfloor › export.video animatic, 0:40

    Here’s the board: nine beats, each held for its exact frames under the new voice, from the tool call that stalls the answer to “Tag your own notes”. Want it animated with sound?

  2. Animate it. Sound effects on the hits, no music, and make the voice loud.

    I’ll lay the nine scenes on the voice at their storyboard frames, put a whoosh, pop or tick on every hit, and bring the voice up front.

    • Used cuttingfloor › timeline.create widescreen, 30 fps
    • Used cuttingfloor › edit.batch nine scenes at their frames
    • Used cuttingfloor › media.import whooshes and pops Claude made
    • Used cuttingfloor › edit.batch 31 effects on the hits
    • Used cuttingfloor › qa.check no gaps, no black frames
    • Used cuttingfloor › audio.set_gain voice up front
    • Used cuttingfloor › audio.set_master loud, no clipping
    • Used cuttingfloor › export.video widescreen, 0:40

    Your 40-second explainer is ready: every beat moves on its word, the voice sits up front and the effects land on the hits. Want a vertical cut for Shorts?

Shown in your agent: Claude, ChatGPT or Kazi. cuttingfloor does the editing and hands back the file; your agent shows it to you.

03

After

Ready to post · 2:11 → 0:40

tip-1-motion-graphics.mp4 · 0:40

What the agent did

  1. Followed two skills: storyboard first, then motion graphics like a pro.
  2. Cut the 2:11 script to 38 seconds, one idea per beat.
  3. Read it in a British voice, with every word timed.
  4. Gave each of nine beats a job (hook, problem, cost, turn, fix, how, payoff, close) and its exact frames.
  5. Built each beat as its own animated scene: things overshoot as they land, arrive one after another, and the camera always moves.
  6. Showed the idea instead of a slide: the answer stops for a tool call, then notes ride the stream between two tags.
  7. Checked every scene frame by frame and fixed five layout slips before the final render.
  8. Laid the scenes on the voice, with whooshes, pops and ticks on the hits and no music.
  9. Checked for gaps and black frames and made it loud and clear.