How to create AI videos from prompts

Start with the job of the video

Before writing a prompt, decide what the clip has to do. A product demo, a short ad, a social hook, and a background visual all need different instructions. Write one plain sentence first: "Create a 10 second vertical video that shows a new skincare bottle on a bathroom counter with soft morning light." That sentence becomes the guardrail for every scene, image, and voiceover choice that follows.

In Eaxy, treat video generation as part of a wider creation workflow. You can draft the idea with writing or chat tools, create supporting visuals with the AI image generator, then move into the AI video generator when the scene is clear enough to animate.

Use a scene-first prompt structure

AI video works best when each prompt describes one shot. Avoid asking for a full commercial, a product launch, three locations, five characters, and a logo reveal in one instruction. Break the video into short scenes, then generate or refine them one by one.

A reliable prompt structure is:

• Subject: the person, product, place, or object the viewer should notice first.
• Scene: where it happens, including lighting, background, and mood.
• Action: what changes during the clip, such as a slow turn, a hand entering frame, or a camera push-in.
• Camera: shot type, angle, lens feel, movement, and framing.
• Style: realistic, cinematic, clean studio, UGC, product ad, animation, or another clear visual direction.
• Format: duration, aspect ratio, and whether it should feel like an ad, explainer, tutorial, or social post.

Example prompt you can adapt

Use this as a starting point, then change the details for your own asset:

Prompt: "Create a 7 second vertical AI video for a social media ad. A matte black wireless speaker sits on a clean desk near a laptop. Soft daylight comes from the left. The camera starts close on the speaker texture, then slowly pulls back as subtle sound waves glow around it. Realistic product photography style, smooth motion, premium tech mood, no text on screen."

This prompt works because it keeps one subject, one setting, one camera move, and one visual idea. It does not ask the model to invent a whole campaign at once. If you need text, captions, or narration, draft those separately so the video prompt can stay focused on the image sequence.

Add source images when consistency matters

If your video needs a specific product, person, room, or visual style, start with a strong source image. A source image gives the AI video tool a visual anchor, which can reduce drift across motion, lighting, and object shape. For product clips, create or upload the clearest still image first, then use the video prompt to describe what should move.

For example, do not prompt "make a video of my product looking premium" if the product design matters. Use a product image, then prompt the scene: "Keep the same bottle shape and label. Add a slow clockwise turn on a white stone surface with soft studio reflections."

Write narration after the visual plan

Voiceover should support the clip instead of fighting it. Once the visual sequence is stable, write the voiceover in short lines. If the video is seven seconds, the narration should usually be one sentence, not a paragraph. Eaxy also supports voice workflows through its AI voice generator, so you can turn concise text into narration after the scene is planned.

Review the result like an editor

When the first generation comes back, check four things before publishing: subject consistency, motion quality, camera clarity, and message fit. If the subject changes shape, simplify the scene or use a stronger source image. If the motion feels random, specify one movement. If the clip is visually good but strategically weak, revise the opening frame or the CTA.

For creators and teams, the practical workflow is simple: plan the shot, generate the asset, write or record supporting copy, and iterate. Create a free Eaxy account to build AI videos from prompts alongside images, writing, voice, transcription, and chat-assisted creative work in one studio.