How we tested this
Reviewed against AI video tool docs in May 2026
We refreshed this hub using current Sora and Runway video-generation guidance, plus the patterns we saw across avatar, animation, and text-to-video tools.
The prompts separate scene generation from scriptwriting, avatar workflows, and motion transfer because each tool family needs different control language.
How to use these prompts
Prompt video like a shot brief. Define the subject, action, camera, duration, scene continuity, lighting, aspect ratio, style, and audio. For image-to-video, describe motion that should happen to the uploaded frame.
Prompt map
Workflows and starter prompts
Use these workflows to choose the right prompt structure for the type of AI video you are creating.
| Job | Prompt focus | Starter prompt | Human check |
|---|---|---|---|
| Text-to-video scene | subject, action, camera, duration | Create an 8-second [aspect ratio] video of [subject] doing [action] in [setting]. Camera: [movement]. Lighting: [lighting]. Mood: [mood]. | Check physics, continuity, hands, faces, and object behavior. |
| Image-to-video | uploaded frame, motion, camera, preserve identity | Animate this image with subtle [motion]. Preserve the subject identity, outfit, and composition. Camera: slow [move]. No scene change. | Reject outputs that distort the original character or product. |
| Ad clip | buyer, product, proof, CTA, brand safety | Create a 15-second ad concept for [offer]. Show [buyer] experiencing [benefit], realistic scene, no fake text, and space for CTA overlay. | Add final claims and CTA in editing, not inside the generated video. |
| B-roll | shot list, mood, matching footage | Create 10 AI video B-roll prompts for a video about [topic]. Each should include subject, action, camera, lighting, and where it fits in the script. | Use only clips that match the narration and do not imply false events. |
| Avatar or explainer | script, presenter, scenes, on-screen text | Create a 2-minute avatar video script about [topic] with 5 scenes, presenter notes, on-screen text, and a final recap. | Check script truthfulness, avatar consent, and language quality. |
Copy this first
The reusable prompt
Text-to-video prompts
Use these for Sora, Veo, Runway, Pika, Kling, and similar scene generators.
Prompt 1
An 8-second cinematic close-up of [subject] in [setting], slow push-in camera, natural light, subtle background motion, realistic physics.
Prompt 2
A 12-second product reveal of [product] on [surface], rotating slowly, soft studio reflections, clean background, no text.
Prompt 3
A vertical social clip showing [action] in 3 fast cuts, bold color, energetic movement, space for captions, strong first 2 seconds.
Prompt 4
A calm educational B-roll shot of [process], overhead camera, clear hands, realistic motion, neutral background.
Image-to-video and motion prompts
When you upload an image, the prompt should control motion without changing the identity.
Prompt 1
Animate this product image with a slow camera orbit and subtle light movement. Preserve product shape, color, and background.
Prompt 2
Animate this character with a gentle head turn and small hand gesture. Keep face, outfit, and proportions consistent.
Prompt 3
Turn this poster into a 6-second motion background. Move only the smoke, light, and particles. Keep the layout stable.
Prompt 4
Create a looping ambient video from this landscape image. Add slow clouds, water movement, and no camera shake.
Script and storyboard prompts
Use these before generating clips so the video has a plan.
Prompt 1
Create a 6-shot storyboard for a 30-second video about [topic]. Include shot, action, camera, narration, and edit note.
Prompt 2
Turn this script into AI video prompts, one prompt per sentence, with matching camera and style notes.
Prompt 3
Create 12 B-roll prompts for this narration. Each prompt should support one line without inventing claims.
Prompt 4
Audit this AI video prompt for vague action, missing camera, identity drift risk, and safety concerns.
What to check before publishing AI video
FAQs
What is the best AI video prompt format?
Use subject, action, setting, camera movement, duration, aspect ratio, lighting, style, audio, and continuity rules. Video prompts need motion and time, not just visual style.
How are image-to-video prompts different?
Image-to-video prompts should describe what moves while preserving the uploaded frame. Tell the model what to keep stable: identity, product shape, logo, outfit, background, or camera framing.
Should I include negative prompts for AI video?
It depends on the tool. Runway's Gen-4 guide says to describe what should happen rather than what should be avoided. In general, positive motion instructions work better than long negative lists.
How long should AI video clips be?
Most generation tools perform best on short clips. Create 5 to 15 second shots, then edit multiple clips together for a longer video.
Can AI video tools create ads?
Yes, but ad scripts and claims need human review. Generate backgrounds, scenes, or product concepts, then add final copy, claims, and CTA in editing.
How do I keep characters consistent?
Use a reference image, stable character description, repeated costume details, and short motions. Regenerate if the face, proportions, or outfit drift.
What should I avoid in AI video prompts?
Avoid non-consensual likenesses, realistic fake news, medical or financial claims without support, protected characters, and prompts that imply real events that did not happen.
What is the main AI video prompting mistake?
The main mistake is writing a still-image prompt. Video needs action, camera movement, timing, continuity, and an edit plan.