Video prompts: camera, action, sound, timestamps

Last updated 13 Oct 2026

An 8-second clip needs its own script: what happens when, what we hear, who speaks and who stays silent.

Course · Stage 5: Consistent AI images & video · Lesson 19/40 · 11 min

Why it matters

Without timestamps the AI often talks over the whole clip or cuts the line short; without sound notes the clip is silent or gets random music.

Without "one continuous shot, no cuts" the AI may stitch several jumpy shots together.

Step by step

1. Open with the frame (9:16 / 16:9), "exactly 8-second single continuous shot, no cuts" and "use the reference image as the start frame".
2. Camera: one move (slow push-in, static, gentle pan).
3. Action: 2–3 things in order, matching the start frame.
4. Sound: specific ambience (ice clinking, water pouring) + music or none.
5. Speech: if the character talks, give timestamps (speaks 0.4–7.4 s, silent at the end) and the line in quotes; if you'll add voice-over later, write "NO SPEECH: nobody talks".

Common mistakes

• No timestamps, so the line gets cut.
• Planning a voice-over but forgetting NO SPEECH, so the clip has stray talking.
• Action that doesn't match the start frame.

Ready-made prompt (ChatGPT / Claude / Gemini)

Write an English video prompt for an 8-second clip, [9:16], start frame: [image description]. Camera: [move]. Action: [2–3 steps]. Ambience: [sounds]. Speech: [the line + timestamps / or NO SPEECH]. Add: single continuous shot, no cuts, no subtitles, no watermark.

Exercise

Write two prompts for the same scene: one where the character speaks, one NO SPEECH for your own voice-over.

Done when: The speaking version has timestamps and the line in quotes; the NO SPEECH version says nobody talks; both have ambience.

Need help? ngocdiep3135@gmail.com