picto.videoAI video prompts: what to write and what to cut
One sentence about what moves. The photo does the rest.
September 2026 · 2 min read
Over the limit? Do this
- Find the one sentence that says what moves. Keep it.
- Delete everything that describes the photo: the room, the light, the lens, the mood, the style.
- Delete every action after the first two. The clip is five seconds.
- Check the counter under the field. Under 2500 and you can generate. Under 100 is normal.
A pasted prompt, trimmed
Before (the first 60 words of a typical paste):
Create a cinematic, ultra-realistic video of the elderly woman in the photograph seated by the sunlit window. The scene should evoke warmth and nostalgia, with soft golden-hour lighting filtering through lace curtains. Shot on 35mm film with subtle grain and anamorphic lens flare. She slowly turns her head toward the camera and smiles gently as dust motes drift...
After:
the woman turns her head to the camera and smiles
Same clip. Nine words.
The recipe
who or what in the photo + does one thing + camera move (optional) + then ... (optional)
Real prompts from the app, as typed:
| Photo | Prompt |
|---|
| A couple | smile and look at each other |
| A dog in the park | pet chases the camera |
| A serum bottle | product drops to the still water creating a natural splash |
| Two people | person on the left must turn and smile |
| A portrait | slight head turn |
Keep this, delete that
| Keep | Delete |
|---|
| The subject, named the way you would point at it: the man on the left, the dog, the bottle | The scene. The model can see the photo |
| One motion, two at most | Style words: cinematic, 35mm, film grain, lens flare, golden hour |
| A camera move in plain words: push-in, pan left, tilt up, hold still | Anything the photo already shows |
| Where the motion ends: "then settles back" | Physics stunts: bouncing, throwing, falling |
Camera words that work
Framing: close-up, medium shot, full body, wide shot. Movement: push-in, dolly-in, pan, tilt, tracking shot. Pace: slow, steady, quick.
No opinion about the camera? Leave it out. A still camera with a moving subject is the safest clip you can ask for.
Why a long AI video prompt fails
A chatbot writes for an image generator: scene, light, lens, mood. The video model already has all of that in your photo, and a description that drifts from the image can make it cut to a different shot. The second problem is time. A clip is five seconds, so "smiles, waves, turns and zooms" gives the model four jobs and time for one. Bouncing balls and thrown objects come out wrong more often than right.
How long an AI video prompt should be
The hard limit is 2500 characters and the app will not send more. The prompts in the table above are two to nine words. A prompt with a subject, a motion, a camera move and an endpoint runs to about 25 words. Past a paragraph you are describing, not directing.
Questions people ask
- Why does the app say my prompt is over the limit?
- The model refuses prompts longer than 2500 characters, so Picto.Video stops you before it charges credits. Delete the description of the scene and keep the movement.
- Do I have to write a prompt at all?
- No. Every lane in the app has styles and presets, and each one is a prompt already written.
- Should the prompt describe the photo?
- No. The model reads the photo. Write what changes. If something must stay fixed, say that.
One photo, one sentence
Upload a photo, type what should move, and watch it happen.
Animate a photo