All of these ran through Grok Imagine, Kling 2.6 and Veo 3.1 before earning a spot. The single rule that outranks everything: narrate the movement, never the picture \u2014 the picture brought its own scene.
Describe the camera, not the subject — instant cinematic feel.
🧑 Portraits & People
Small, natural motions keep faces believable.
🐕 Pets & Animals
The internet’s favorite — keep motion energetic but short.
📦 Products & E-commerce
Slow orbits and lighting changes sell products.
🏞️ Landscapes & Nature
Animate the environment, keep the horizon stable.
🖼️ Old Photos & Nostalgia
Bring family pictures back to life with gentle motion.
🔥 Memes & Social
Punchy, exaggerated motion wins feeds.
✨ Fantasy & Sci-Fi
Big effects work best with one clear idea per clip.
🎛️ Style Modifiers
Append these to any prompt above for a different look.
How to write your own prompt
The formula that works on every model: [subject action] + [camera behavior] + [mood/style]. Example: "the couple dances slowly (action) while the camera circles them (camera), warm candlelight, film look (style)". One action per clip. If you need two things to happen, generate two clips.
FAQ
What makes a good image-to-MP4 prompt?
Say what moves, not what exists — the frame already holds the scene. A workable prompt names one camera instruction (push-in, orbit, pan) or one subject action (turns, runs, smiles), optionally a style tag. One idea per clip is the ceiling.
How long should a video prompt be?
One to two sentences across Grok Imagine, Kling 2.6 and Veo 3.1. Pile on more and the model averages your directions into porridge.
Do these prompts work on any AI video model?
They are deliberately engine-agnostic. Verified here on Grok Imagine, Kling 2.6 and Veo 3.1; the same sentences behave fine on Runway, Luma and Pika.
Should I describe the scene in my prompt?
Exclusively the delta. The photograph already fixed the scene — re-describing it invites the model to repaint and drift. Motion and mood only.
Why did my generation ignore the prompt?
Typically a prompt fighting itself — two camera moves at once — or an adjective pile, or motion the frame cannot support. Cut to one action and re-run; failed attempts here refund their own credits.
Paste a prompt, get a video
transparent pay-per-use pricing from your first clip.