To prompt AI video well, write it like a shot list, not a poem. Name the subject, describe the action in the order it happens (who moves, which part, which direction, how it ends), then add camera, sound, and a one-word mood. Literal, physical language beats clever description.
Practice in VidwurtA clear sequence of actions does more than any stack of words like “epic” or “cinematic.”
When you animate a picture, the picture sets the look. Your words should set the motion.
A push-in or an orbit, not both. Split complex moves into separate clips.
One mood word at the end reinforces the clip. At the start it pushes the action aside.
Most video models read a prompt from the top and give the first words the most weight. So lead with what must happen, and add the finish after. This order works for text to video and image to video on Wurt.app:
A text to video prompt has to build the whole scene, so describe the subject, setting, and light before the action. An image to video prompt is different: the picture already fixes the face, clothes, room, and style, so re-describing them wastes words and can confuse the model. Spend almost the whole prompt on what happens next.
If the picture is missing something the action depends on, such as a door the character should walk through, fix the picture first in Picwurt and then animate it. If you just want to see how a still moves, Quick Animate does it in one tap with no prompt at all.
Weak: “Epic cinematic dragon, amazing, dramatic, masterpiece.”
Strong: “A green dragon glides low over a pine forest from left to right, beats its wings twice, then climbs steeply and disappears into the clouds. The camera tracks alongside at treetop height. Wind roars and each wing beat thumps. Epic.”
The strong version tells the model what moves, in which direction, and how the clip ends. “Epic” still appears, but only once, at the end.
Weak: “A beautiful woman in a red dress standing in a garden with roses, smiling, soft light.”
Strong: “She lifts the hem of her dress with her right hand, takes two steps forward on the stone path, and looks back over her shoulder at the camera. Slow push-in. Soft birdsong.”
The picture already shows the dress, the garden, and the light. The rewrite spends every word on motion.
Weak: “Make my sneaker look cool and dynamic.”
Strong: “The sneaker turns one full rotation clockwise on the white pedestal. Halfway through, the camera lowers to ground level. The clip ends on the side profile with the logo facing the camera. Clean studio light.”
“Cool” and “dynamic” mean different things to every model. A rotation, a camera change, and a final frame do not.
Use the exact film term when you know it. A precise term anchors the shot better than a long description of the same idea.
Wurt, Sqwurt, Horse, Seedance 2.0, and Seedance 2.5 can add sound to the clip (see AI video generator with sound). When they do, write sound in the same plain way as motion. For speech, put the line in quotes and say how it is delivered: “she whispers, ‘come here,’ slowly and warmly.” For a laugh, a gasp, or a door slam, just name the sound. If you want no sound, say so. GIFs are silent, so leave sound out of GIF prompts.
Want to go deeper on still images first? The prompt engineering guide covers picture prompts, and the Learn hub lists every guide. For platform-specific pacing, see AI videos for TikTok.
Put the subject first, then the action in the order it happens, then the end state, camera, sound, and a short mood word. Physical, literal instructions work better than poetic descriptions.
Long enough to describe every action that matters. For a 5 to 10 second clip that is usually two to five sentences. Cut repeated adjectives and filler, but keep each concrete detail that changes the result.
Only briefly. The start image already shows the subject, setting, and style, so spend your words on what moves next and how the clip ends. Use short labels such as 'the man on the left' to say who moves.
Models give the most weight to the start of the prompt and to concrete verbs. Put the main action first, use one camera move, describe what you want instead of what you don't want, and remove vague mood words that compete with the action.
Yes. Wurt AI, the chat helper in Wurt.app, can turn a rough idea into a full video prompt or fix one that missed, and the first 5 messages are free. If your flow shows High Fidelity, that optional setting also refines the prompt before it goes to the model.
Write one prompt with the subject, the action in order, and a clear ending, then compare it with your old version. Eligible new accounts can get free starting credits.
Practice in Vidwurt