VIDEO TO PROMPT
Upload a short clip or a single image and get a video prompt written for Sora, Veo, Kling or Runway: camera, motion, light and timing.
Add a short clip and four frames are sampled in your browser — the video itself never leaves your device. Or add a single image to get an image-to-video prompt.
“Slow push-in on a woman holding a transparent umbrella on a rain-washed Kyoto platform at blue hour. Rain streaks the frame; she turns her head toward an arriving train whose warm headlights sweep across wet paving. 0–4s: static medium shot, rain and steam drifting. 4–8s: the camera eases forward as the train enters from the right. Audio: soft rain, distant announcement chime, train brakes.”
EXAMPLE WORKFLOW
Stills and clips become video prompts that say what moves, where the camera goes, and how long the shot lasts.
Slow push-in on a woman in a moss-green wool coat holding a transparent umbrella on a rain-washed Kyoto platform at blue hour. 0–3s: static medium shot, rain streaking the frame, coral lanterns reflected in wet paving. 3–8s: the camera eases forward as a train enters from the right and its warm headlights sweep across her face; she turns toward it. Cool ambient light against warm practicals, 35mm film texture, quiet cinematic mood. Audio: steady rain, a distant platform chime, train brakes hissing.
IMAGE TO VIDEO PROMPT · SORA 2
The image supplies subject, light and mood. The prompt adds the camera move, the action and the timing Sora needs.
Slow push-in on a woman in a moss-green wool coat holding a transparent umbrella on a rain-washed Kyoto platform at blue hour. 0–3s: static medium shot, rain streaking the frame, coral lanterns reflected in wet paving. 3–8s: the camera eases forward as a train enters from the right and its warm headlights sweep across her face; she turns toward it. Cool ambient light against warm practicals, 35mm film texture, quiet cinematic mood. Audio: steady rain, a distant platform chime, train brakes hissing.
A slow 180-degree orbit around a translucent amber perfume bottle on a wet cobalt-blue plinth, macro lens feel with shallow focus. As the camera circles, a softbox highlight glides across the glass and the ivory cap; a pale paper ribbon lifts in a gentle breeze and settles; faint ripples spread across the plinth. Deep cobalt background, glass and satin textures, quiet luxury. Sound: a low airy studio tone and one soft water drip.
IMAGE TO VIDEO PROMPT · VEO 3
Product stills become short hero clips when the prompt names the orbit, the light behaviour and the one small motion in the scene.
A slow 180-degree orbit around a translucent amber perfume bottle on a wet cobalt-blue plinth, macro lens feel with shallow focus. As the camera circles, a softbox highlight glides across the glass and the ivory cap; a pale paper ribbon lifts in a gentle breeze and settles; faint ripples spread across the plinth. Deep cobalt background, glass and satin textures, quiet luxury. Sound: a low airy studio tone and one soft water drip.
A giant tortoise carrying a small brass-and-wood observatory walks slowly through a moonlit wildflower meadow; the flowers sway as it passes and warm light flickers in the observatory windows. Wide shot, slow tracking camera moving left to right at the tortoise's pace. Hand-painted gouache storybook style, indigo sky with drifting clouds, silver moonlight, calm and wondrous.
VIDEO TO PROMPT · KLING
From four sampled frames the tool reads the subject's action and the camera move, then writes them in Kling's compact order.
A giant tortoise carrying a small brass-and-wood observatory walks slowly through a moonlit wildflower meadow; the flowers sway as it passes and warm light flickers in the observatory windows. Wide shot, slow tracking camera moving left to right at the tortoise's pace. Hand-painted gouache storybook style, indigo sky with drifting clouds, silver moonlight, calm and wondrous.
WRITE MOTION, NOT A PICTURE
An image prompt describes a frozen frame. A video prompt has to describe change: what the subject does over the next eight seconds, how the camera moves, how the light behaves, and when each beat happens. This tool reads that change out of a clip or infers it from a single image, then writes it in the syntax your video model responds to.
Frames are sampled in time order, so the tool can compare them: a subject that shrinks between frames implies a pull-back, a horizon that tilts implies a handheld move, a light source that sweeps across a face implies a passing car or a turning head. It names the shot size, camera height and angle, the subject’s action and how it progresses, the setting, time of day and weather, the lighting direction and quality, and the palette and style signature.
It then writes the prompt as motion. Every sentence says what moves, including the camera. The style is preserved rather than upgraded: grainy phone footage stays grainy phone footage unless you ask for something else, because that is usually the point of extracting the prompt.
With a single image there is no motion to read, so the tool infers a plausible one from the picture: a portrait gets a slow push-in and a small gesture, a product gets an orbit and one moving element, a landscape gets a drift and weather. Your direction overrides all of it. “She looks up and smiles” or “static camera, only the water moves” are the sentences that make an image-to-video prompt yours.
The prompt keeps what the image already decided. Subject, framing, light and palette are described so the video model’s first frame matches your picture, which is what Sora, Veo, Kling and Runway image-to-video modes need to stay faithful to the reference.
The same idea has to be written five ways. Sora 2 and Veo 3 generate sound, so their prompts end with an audio line; Kling weights the subject and a single clear motion; Runway wants the camera move first, in its own words. Choose the model in the tool and the prompt follows these rules.
| Video model | How the prompt is written | Audio | Length |
|---|---|---|---|
| Sora 2 | Plain prose; a short shot list with timings when there are several beats | Yes, native audio: end with a sound or dialogue line | Under ~180 words |
| Veo 3 | One cinematic paragraph: shot, action, environment, light, style | Yes: add a sound-design sentence, dialogue in quotes | Under ~160 words |
| Kling | Subject → motion → scene → camera → light, one clear primary motion | No | 60–110 words |
| Runway Gen-4 | Camera motion first (push in, orbit, tracking, crane), then subject motion | No | Under ~80 words |
| General | Cross-model prose: shot and camera, subject motion, setting, style, duration | Optional | 90–160 words |
Creators and marketers turn a product photo or a campaign still into a short hero clip without writing camera language from scratch. Filmmakers and editors extract the prompt from a reference shot to recreate its movement and light in Sora or Veo. Social teams reproduce the pacing of a clip that performed well with a new subject. Prompt engineers use the model-specific formats as a starting point and compare how Sora 2, Veo 3 and Kling interpret the same shot.
Your video is not uploaded. Four frames are sampled by your browser at reduced size and only those frames are analysed; nothing you submit is added to the Gallery or shown to anyone else. Clips are limited to 60 seconds and 60 MB, which is enough for a single shot and keeps analysis fast. A run costs 2 credits, twice the price of the image tools, because it reads several frames and writes a longer, timed prompt.
Need the prompt behind a still image instead? Use Image to Prompt. Starting from words? Text to Prompt expands them, and the Image Prompt Gallery shows finished prompts by style.
COMMON QUESTIONS
It reads a short video, or a still image, and writes the text prompt that would recreate that shot in an AI video generator. The prompt describes what moves: the subject's action, the camera movement, the light, the pacing and the duration, in the format Sora, Veo, Kling or Runway reads best.
Upload an MP4, WebM or MOV clip of up to 60 seconds. Four frames are sampled across the clip in your browser and sent for analysis; the video file itself never leaves your device. Choose the video model and target duration, add optional direction, and click Generate. Trim long videos to the single shot you want a prompt for.
Yes. Add one image and the tool writes an image-to-video prompt: it keeps the subject, framing, light and style of the picture and adds the motion, camera move and timing a video model needs to animate it. This is the right input for Sora, Veo, Kling and Runway image-to-video modes.
Yes. Choose Sora 2 and the prompt is written as plain prose with a short shot list when the motion has several beats, each with timing in seconds, and a closing line for audio, because Sora 2 generates synchronized sound.
Each has its own option. Veo 3 receives one cinematic paragraph with an audio line; Kling receives a compact prompt ordered subject, motion, scene, camera, light; Runway Gen-4 receives a short prompt that leads with camera motion in Runway's own vocabulary. Choose General for a prompt that works across models.
No tool can read the hidden prompt out of a finished video. What it does is describe the visible motion, camera, light and style precisely enough that running the prompt again produces the same kind of shot.
You can try it without an account within the daily free allowance. Each run costs 2 credits on a plan, because it reads several frames and writes a longer prompt than the image tools.
MP4, WebM and MOV up to 60 MB and 60 seconds, plus JPG, PNG and WEBP images. Because frames are read in your browser, very old browsers may not support every codec; if a clip fails to load, export it as MP4 (H.264) and try again.