How to make a video with AI

Step 1: decide what the clip is for

One clip should do one job — stop the scroll, show a product, set a mood, open a longer video. Knowing the job tells you the shape and length before you write a word, and it stops you asking a single render to carry a whole story.

Step 2: write the shot, not the idea

"A video about coffee" gives the model nothing to aim at. "A cup of black coffee on a wooden bench, steam rising, slow push in, warm morning light through a window" gives it a subject, an action, a camera move and a light source. That is the whole trick.

• Subject — one clear thing in frame
• Action — one thing happening, not three
• Camera — push in, orbit, tilt up, whip pan
• Light — morning window, neon, overcast, backlit
• Look — 35mm grain, glossy product, anime, 3D render

Step 3: pick length and shape

Five to ten seconds is where generated video looks best; longer takes drift. Choose 9:16 for TikTok, Reels and Shorts, and 16:9 for YouTube, websites and ads. Generate in the shape you will publish in rather than cropping afterwards.

Step 4: or skip the writing and use a photo

If you already have an image of the subject — a product, a portrait, artwork — upload it instead. Expivid scans it, writes the motion for you and uses the photo as the first frame, so the subject stays recognisable while the camera moves.

Step 5: judge the render, then change one thing

If the clip is soft, you probably asked for too much motion at once: cut to one action. If the subject warps, name it more plainly. If the mood is wrong, change the light, not the subject. Changing one variable per attempt is much faster than rewriting the whole prompt.

Step 6: finish it elsewhere

Generated clips arrive silent and without text. Add captions, music or a voiceover in whatever editor you already use, and cut two or three clips together if you need something longer than ten seconds.