NicheLens8-second shots and seamless chaining
Last updated 13 Oct 2026
Each clip is one continuous shot: one action, one camera move. The last frame of one is the first frame of the next.
Course · Stage 5: Consistent AI images & video · Lesson 14/24 · 9 min
Why it matters
Most AI video tools generate about 8 seconds at a time. Cramming three actions into 8 seconds warps the image and deforms characters.
Chaining (using the last frame as the next start frame) makes a long video look like one real take, with no jumps.
Step by step
- Split the script into 8-second scenes, one main action each.
- Give each scene one camera move: push in, pull out, pan, locked off.
- The scene image is the FIRST moment of the clip, not the middle.
- For seamless flow: scrub to the end of the previous clip, save the frame, use it as the next start frame (Flow / Kling have Extend).
- Vary shot size between scenes (wide → medium → close); never 3 in a row the same.
Common mistakes
- Three actions in one clip.
- The image shows the middle of the clip.
- Ten scenes from the same angle.
Ready-made prompt (ChatGPT / Claude / Gemini)
Split this script into 8-second scenes: [paste]. For each: one main action, one camera move, shot size (wide / medium / close — never 3 in a row the same), the first moment to generate as an image, and the last moment to chain into the next scene.
Exercise
Split a 48-second script into 6 scenes using these rules.
Done when: 6 scenes, each with exactly one action and one camera move, alternating shot sizes.
Check yourself
Get all 3 right to mark this lesson complete.
1. How many main actions per 8-second clip?
2. Which moment does the image represent?
3. How do you chain scenes?
Need help? ngocdiep3135@gmail.com