Editing & refinement

How to change a finished video — conversational edits, scene-level regeneration, clip extension and editing, transitions, and versioning.

Last updated July 2, 2026

Generation is the first draft. Everything after is editing — and every edit is scoped, so changing one thing never regenerates everything else.

Edit by talking

Open any session and describe the change. Wavemaker understands your original brief, the full composition state, and the conversation so far:

“The hook is weak — open on the product close-up instead.” “Scene 3’s lighting is too dark.” “Swap scenes 2 and 3.” “Cut the video to 25 seconds.” “The last two scenes don’t match what I asked for — redo them from the brief.”

Each refinement saves a new version; earlier versions remain available.

What’s editable, precisely

You sayWhat happens
”Regenerate scene 2’s image”New still for that scene, quality-reviewed, re-animated
”Re-animate scene 5, keep the image”The existing frame is kept; only the motion re-rolls
”Extend scene 4 by 3 seconds”The clip is extended seamlessly from its last frame
”Remove the flicker in scene 6”A targeted video edit — cheaper than regeneration, keeps the take
”Replace the mug with the blue one”Element-level video edit — one object swaps, the rest of the clip stays
”Slower transitions” / “add a light-leak between 2 and 3”Transition re-planning, per-cut or global
”Change the headline on the end card”Text/overlay edit — no media regenerated
”New narrator” / “requote scene 1’s line”Audio-layer changes only

Smart-edit guarantees

  • Identity survives edits. Subject references, the designed narrator voice, and each character’s voice persist across refinements — the “same person, new scene” problem is handled.
  • Baked dialogue is preserved. Re-animating a scene where a character speaks keeps the visual and re-bakes the same line — the words aren’t lost or changed unless you change them.
  • Broadcast masters stay exact. Editing a :30 spot re-fits the composition to exactly 30.00s.
  • Pixels untouched, checks skipped. A text-only or audio-only edit doesn’t re-run visual quality gates on unchanged footage — it’s fast.

The step-by-step mode

Prefer to approve as you go? Step-by-step mode pauses at phase boundaries (after planning, after the storyboard) and waits for your go-ahead — useful for expensive, high-stakes productions.

When to edit vs. regenerate

  • Edit (fix, extend, swap an element) when the take is right but a detail is wrong — it’s faster and cheaper.
  • Regenerate the scene when the composition itself is wrong.
  • New generation when the concept changed — start a fresh session and reference what you liked.

For visual, node-level control of all of this, see Canvas.