The workflow
Pick a clip with room on both sides
Use one continuous shot from 3.5 to 60 seconds. The new scene has to start at least 1.7 seconds in and leave at least 1.5 seconds of your video after the point where it picks up again. A longer clip is trimmed to the 60-second window you choose before it is sent.
Choose where the scene goes
Set the cut point where your video stops and the point where it resumes. The footage between them is replaced. The model sees your shot on both sides of the gap, so it can start from your frame and blend back into it. A moment with steady light and framing makes the cleanest transition.
Describe what happens
Write the direction in plain words, or start from an idea: your product, someone new, a surprise, the weather, or a close-up. To bring in a product or a person, add a photo and call it Image 1 in the direction. Say who does what, from where, and how it ends, since the scene has to hand back to your footage.
Check the quote, then watch both cuts
Pick the scene length and read the price; only the new seconds are billed. After it finishes, watch the cut in and the cut out at normal speed. Check faces, hands, and the subject’s position at the hand-back, where a mismatch is easiest to notice.
Recorded Made demonstration
One walk, with a dog that wasn’t there
This real Scene Insert result was generated on September 29, 2026 from a Made-original street clip. The scene starts 3.5 seconds in, replaces two seconds of the walk with a five-second scene, and returns to the original at 5.5 seconds. The finished video is about 13 seconds. The confetti version uses the same cut with a different direction.
MiniMax H3 Max · Generated · View the example record and price basis (JSON)
Prompts you can inspect and adapt
Recorded generation prompt
The exact direction used for the dog scene
Her golden retriever sprints into frame from the side and leaps up into her arms; she catches him, laughs and spins around once.Short and physical: who enters, from where, what happens, and how it ends. The model generates its own sound for the new scene.
Recorded generation prompt
The direction for the confetti version
Two huge confetti cannons blast from both sides of the frame, filling the air with a thick shower of gold and white confetti that rains down all around her while she throws her arms up and laughs.Same clip and same cut points. Only the direction changed.
Prompt recipe to adapt
Adapt it for a product shot
Cut to a close-up of the product in Image 1, placed naturally in this scene with the same light and setting. Keep its shape, colors, label and details exactly as in Image 1.This is the tool’s built-in “Your product” idea. Add the product photo first; label text can still drift, so check it in the result.
How a new scene is priced
By the seconds of new scene you add. Your own footage before and after the cut isn’t charged. The total appears on Generate before you run it.
Only the new scene is billed, per second at 768p, so a 10-second scene costs twice a 5-second one.
Price source: Public Scene Insert catalog · Checked
The footage before and after the cut is your own video and isn’t charged, whatever its length. Up to two photos don’t change the price. Each new attempt is priced again, so test a five-second scene before trying a longer one.
What to review before you use the result
| What you see | What to try |
|---|---|
| The cut back to your video jumps | Move the resume point to a calmer moment where the subject is in a similar spot, or end the direction with the subject returning to where they were. |
| The tool won’t accept the cut points | Start the scene at least 1.7 seconds in and keep at least 1.5 seconds of your video after the resume point. |
| The scene was turned down | Directions and uploads are checked for safety, and failed runs aren’t charged. Rewrite the direction, and use footage of adults who agreed to it. |
| A product’s label changes | Use a sharp product photo and name it as Image 1. For exact branding, treat the result as a draft and fix text in your editor. |
Work with Claude and Made
In Claude, connect Made and describe the clip, the two cut points, and what should happen. Ask the assistant to inspect the catalog for the insert model and quote the scene length first.
Conversation starter
Ask your assistant
Use Made to add a new five-second scene to my attached video. Stop my video at 3.5 seconds and pick it up again at 5.5 seconds. In the new scene, my dog runs in from the side and jumps into my arms. Check the current catalog for MiniMax H3 Max insert, show me the credit quote, then generate one result and return the video link.If the connected client doesn’t expose the insert model or the upload path, open Scene Insert with the same clip. Don’t substitute a text-to-video model, which makes a separate clip instead of one that cuts in and out of your footage.
Questions about this workflow
Does my video keep its sound?
Yes. The parts before and after the new scene keep their original audio, and the new scene gets sound generated to match.
Can I put my product or a person in the scene?
Yes. Add up to two photos and refer to them as Image 1 and Image 2 in your direction. Use photos of people who agreed to it.
How long does it take?
Usually one to four minutes. You can leave the page; the finished video is saved to Assets in your Made account.