Turn raw footage into high-end motion graphics with Claude: the full step-by-step
No more dragging clips along a timeline, no more getting stuck because you can’t build the motion graphic you pictured. This one takes my AI editing workflow apart completely: you direct the AI in plain language, and a raw, stumbling talking-head take comes out the other side as a finished cut with real motion design on it.
A lot of people assume AI-generated video always looks cheap. The tools are only the skeleton — what sets the ceiling on a finished piece is what I call taste feeding: the references you hand the model. This walks through the architecture I actually run, with Claude Code as the director tying HyperFrames, ffmpeg, and Whisper together, so you can build your own automated pipeline without writing code by hand.
What the video covers
- The before-and-after — the raw take against the finished cut, with no manual editing in between.
- Why AI video all looks the same, and the way out — how taste feeding gets Claude to produce motion graphics that actually look considered.
- Installing VSCode — the workspace the whole pipeline lives in.
- HyperFrames or Remotion? — what each one is for, and which I picked.
- Installing Whisper — the speech-to-text step that makes transcript-accurate cutting possible.
- The hands-on build — the full run-through, step by step, nothing skipped.
- My private motion-graphics inspiration library — the reference sites I pull from: the HyperFrames catalog, 21st.dev, Envato Elements, and godly.design.
- Iterating, hands-on — how to push a first pass toward the look you wanted, instead of settling for whatever came out.
Where to start
If you haven’t set up the underlying pipeline yet, start with auto-editing video with Claude Code — this video builds the motion-design layer on top of it. For the generation side of the same craft, see AI storyboards for short video.