Over the limit? Do this

  1. Find the one sentence that says what moves. Keep it.
  2. Delete everything that describes the photo: the room, the light, the lens, the mood, the style.
  3. Delete every action after the first two. The clip is five seconds.
  4. 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:

PhotoPrompt
A couplesmile and look at each other
A dog in the parkpet chases the camera
A serum bottleproduct drops to the still water creating a natural splash
Two peopleperson on the left must turn and smile
A portraitslight head turn

Keep this, delete that

KeepDelete
The subject, named the way you would point at it: the man on the left, the dog, the bottleThe scene. The model can see the photo
One motion, two at mostStyle words: cinematic, 35mm, film grain, lens flare, golden hour
A camera move in plain words: push-in, pan left, tilt up, hold stillAnything 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