An idea for a still image
Describe the subject, setting, framing, and light. Pick an image model before generating.
Open the image creatorWrite · review · refine
Grok Imagine Prompt Guide for Images and Video
Choose the right task, write a clear instruction, and learn what to adjust. Work through image composition, photo edits, video motion, and community prompt breakdowns.
Start with the input you have
A photo you want to change
Upload your photo, identify the edit, and name the important features to preserve.
Open image-to-imageAn idea for a video
Establish the scene, choose the main action, and specify a camera view. Set duration and resolution in the tool.
Open the video creatorA still image you want to animate
Attach the image in a compatible video tool. Describe movement that fits the people and objects already visible.
Explore Grok Imagine 1.5Check the selected model, required uploads, and displayed credit estimate before submitting. Writing a duration or resolution in a prompt does not replace the form settings. Review credits
Make the image instruction specific
Build the scene in layers
Name the subject first, then its surroundings, viewpoint, and light. Add a visual style if it matters. For example, “a red bicycle” becomes “a red bicycle outside a bookshop, side view, morning light from the right.” Each addition answers a different visual question.
Choose details with a purpose
For a portrait, describe the expression and framing. For a product, identify the surface, material, and lighting direction. Use details that determine the picture rather than adding a long list of quality adjectives.
Define the edit boundary
An uploaded photo supplies a starting composition. Identify the object or region to change, the change itself, and the features to keep. A color change and a background replacement can be tried separately so each result is easier to assess.
Review the parts you wanted preserved
For the bicycle example, ask for a green frame while retaining wheel positions and the storefront. Check whether only the intended area changed. Inspect lettering, faces, and small details before using the output.
Give the subject and camera separate jobs
Subject action
Choose an observable change: a shopkeeper turns a sign or a runner crosses a line. Describe the speed separately. For several beats, write them in order and check whether the clip has enough time; split a longer story into separate shots when needed.
Camera direction
A person approaching a fixed camera is different from a camera approaching a stationary person. Specify the starting view and one camera behavior. Avoid asking for a locked shot and an orbit at the same time.
Image-to-video
Refer to the people and objects actually present in the uploaded image. Concentrate on what moves and keep the setting consistent unless a transformation is intentional. Do not contradict the source image with a different subject or location.
Text-to-video
Supply the scene as well as the action. Character appearance, environmental details, mood, and lighting can establish context; prioritize the details that matter to the shot instead of describing every feature equally.
A small camera vocabulary
- Static shot
- Hold the viewpoint while the subject moves.
- Slow push-in
- Move closer to emphasize a detail or expression.
- Tracking shot
- Move alongside the subject.
- Pan
- Turn the view to reveal another part of the scene.
Sound and continuation
When the selected model supports audio, describe a sound that accompanies the action: a sign creaking, footsteps, or a short spoken line. A separate audio sentence makes the brief easier to read; it does not add audio support to a model.
For an extension, inspect the source clip’s ending and describe the next action. Avoid restarting the story or assuming that every detail will stay identical. Review source requirements and credits in the tool.
Open video extensionFour starting prompts to make your own
These are writing exercises, not recorded generation results. Recreate transfers text only. Review the model and mode, replace the subject as needed, and attach your own image for editing or animation. Generation starts only when you submit in the tool.
Image: composition and light
Choose an image model and text-to-image mode.
A red bicycle leaning against a pale stone wall outside a small bookshop, side view, morning light from the right, soft shadows on the pavement, muted colors, documentary photograph.
The bicycle is the subject. The wall provides separation. Side view fixes the viewpoint, while the light direction guides the shadows.
Image edit: change and preserve
Choose an image editing model and attach your own photo.
Change the bicycle frame to forest green. Keep the wheel positions, handlebar shape, bookshop signs, background, and lighting unchanged.
The edit targets frame color. The preservation list identifies details to inspect after generation; it does not guarantee they will remain identical.
Video: establish the scene
Choose a video model and text-to-video mode.
Wide shot of a red bicycle outside a quiet bookshop in morning light. An adult shopkeeper steps out and turns the hanging sign. The camera stays still; a nearby leaf moves in a light breeze.
Without an input image, establish the setting and action. One person and one main action give you a clear result to review.
Image to video: add movement
Choose an image-capable video model and attach a matching still image.
The hanging sign sways gently once and settles. A few leaves drift across the pavement. The camera remains still and the bicycle stays in place.
The uploaded image supplies the appearance. Describe changes within that frame rather than inventing a new setting.
Grok Imagine 1.5 Prompt Tests and Results
These six examples were tested by the Grok Imagine AI site operator using Grok Imagine 1.5. Explore the original prompts, recorded settings, and resulting videos, with source images included for the image-to-video tests.
Identify the subject
The short instruction identifies a person but leaves the movement open. When adapting it, name a visible action rather than assuming that an appearance description will produce one.
Original prompt
The old man wears glasses
Model: Grok Imagine 1.5 · Recorded settings: Fixed lens, 720p, 10s
Attach your own matching image in the creator; Recreate transfers text only.
Separate focus from action
A focus change and a character action are different instructions. Specify who stays visible, where attention moves, and which person acts.
Original prompt
The camera focuses on the teacher in the background, the girl in the foreground becomes blurred, and the teacher curses very angrily
Model: Grok Imagine 1.5 · Recorded settings: Unfixed lens, 720p, 5s
Attach your own matching image in the creator; Recreate transfers text only.
Describe a shot change
This prompt requests both an activity and a cut. Decide whether a cut is essential; a single continuous shot is easier to inspect for consistency.
Original prompt
Kittens and puppies eat cat food.Shot Switch. Close-up cat food is distinct.
Model: Grok Imagine 1.5 · Recorded settings: Unfixed lens, 720p, 5s
Attach your own matching image in the creator; Recreate transfers text only.
Control the pace
Speed words and a follow shot describe different things. Choose the pace of the athlete separately from the movement of the camera.
Original prompt
An athlete in a professional tracksuit. Legs alternate rapidly, arms swing powerfully, sprinting with all his strength on the field. After crossing the finish line, the audience erupts in cheers. The screen uses a follow-shot perspective to fully showcase the details of the athlete crossing the finish line.
Model: Grok Imagine 1.5 · Recorded settings: Unfixed lens, 720p, 5s
Attach your own matching image in the creator; Recreate transfers text only.
Build a scene from text
Without a source image, the prompt supplies the person, lighting, foreground, and camera move. Keep the details that determine the shot and remove details that do not matter to your goal.
Original prompt
Portrait photography with a sense of design, psychedelic cold light blue tones, butterfly light, close-up shot of a young white woman. She has high-level short black hair, raised eyebrows on the right, thick eyelashes, high nose bridge, biting red lips, and staring disdainfully at the camera. The camera pulls back, and the foreground is broken glass in the air, which blocks part of the woman's face.
Model: Grok Imagine 1.5 · Recorded settings: Unfixed lens, 720p, ratio 16:9, 5s
Balance detail and motion
A dense portrait description competes with the requested gesture. For an adaptation, keep the defining appearance and choose one gesture to emphasize.
Original prompt
In the winter valley where the snow is falling, a bright and beautiful young white woman in a close-up shot turns her head sideways to look at the camera. She has long black wavy hair, a pointed chin, bushy raised eyebrows, deep eye sockets, red eyes, dark eyeshadow and upturned eyeliner, straight nose, thick lips, very bright red lips, clear jaw line, very long nails, red manicure. The woman is wearing a black robe, wearing a hat to cover her eyebrows, and the collar is slightly open with clear collarbone. Her eyes are fixed on the camera, and her eyes are very attractive. The background is green vegetation covered with thick snow, and snowflakes fall in the air. The camera slightly surrounds the woman to the left, and the woman raises her right hand on her chin, looking at the camera with a charming smile.
Model: Grok Imagine 1.5 · Recorded settings: Unfixed lens, 720p, scale 9:16, 5s
Grok Imagine community prompts, explained
Read each original prompt beside its linked community video. The notes explain the wording; suggested changes have not been tested as controlled comparisons. Original model settings, charges, and generation times are not supplied by these records.
Community example by Alicia McNatt. Original source
Browse the full prompt librarySeparate subject movement from camera movement
Read the original prompt
Animate this still, 8 seconds, smooth cinematic loop. The Asian man holds the open black Zippo steady, thumb resting on the lid. The small red panda in the purple BAD RUDY beanie sits on the flip cap, mouth open, breathing a short controlled burst of orange fire that flickers and dies back to a small flame on the lighter wick. His glowing orange eyes blink once. Embers drift. The man’s smirk shifts a fraction, one brow lifts, then settles. Forearm tattoos stay still. Purple circuit board behind him pulses once, slow. No camera move. No extra characters. Sound: lighter flint click, soft flame whoosh, one quiet exhale.
How the wording is structured
The prompt names a short sequence of actions, then says “No camera move” and “No extra characters.” These are separate instructions: one controls the subject, another fixes the viewpoint, and another limits additions. Its sound cues are also written separately.
Adaptation to try
For your own scene, keep one main action and specify whether the camera stays still. If starting from an image, supply your own reference; copying “Animate this still” alone does not provide the original image.
Community example by GayBar. Original source
Browse the full prompt libraryGive the scene a concrete goal and mood
Read the original prompt
Make a video of him looking for dry wooden sticks for his home. Trying to stay warm in the winter. He seems to be talking to himself. Worried he cannot find enough wood for his fireplace. Cinematic. #gay
How the wording is structured
Looking for dry sticks supplies an action and a purpose. Winter supplies the setting, while worry about the fireplace supplies an emotional cue. “Cinematic” is a broad style word; the other details give the model a more specific scene.
Adaptation to try
Replace “him” with a description of your subject when no reference is supplied. Start with one action, such as picking up a branch, and add a specific light source instead of relying only on “cinematic.”
Community example by Ami A. Original source
Browse the full prompt libraryDescribe composition and material separately
Read the original prompt
A tight blue-and-white striped Lacoste polo shirt laid flat as a luxury fashion still life on white sheets, sharp product photo. Use the exact style, fabric texture, collar, buttons, and green crocodile embroidery logo from space ai x
How the wording is structured
“Laid flat” and “on white sheets” describe the arrangement. Stripes, texture, collar, and buttons describe visual details. The request for exact branding is an instruction in the source prompt, not evidence that the output reproduces every logo accurately.
Adaptation to try
Use an item you have rights to depict, specify the surface and framing, and inspect lettering and logos in the output. This community record is inspiration, not a verified product-editing before-and-after test.
Community example by Hachiko - $HACHI. Original source
Browse the full prompt libraryTurn a long story into a manageable shot
Read the original prompt
Realistic cartoon style video of Hachiko’s true 1920s Tokyo life: the Akita puppy born in 1923 in Ōdate, adopted by Professor Ueno in 1924, their daily walks to Shibuya Station, Ueno’s sudden death in 1925 from cerebral haemorrhage, and Hachiko’s real, faithful wait every day for nearly 10 years until his passing in 1935 amid crowds and seasons.
How the wording is structured
This prompt compresses a life story, several dates, changing crowds, and seasons into one brief. It illustrates how a detailed narrative can contain many different actions; it does not establish that each event appears faithfully in the result.
Adaptation to try
For a short clip, isolate one beat, such as the dog waiting beside the station entrance. Keep the longer chronology as a storyboard for separate shots rather than expecting one request to show every event.
Choose one thing to change next
The subject is unclear
Replace broad wording with a visible subject and a location in the frame. Remove competing subjects if they are not important.
The motion is too busy
Keep one main action and name its speed. If the scene needs several shots, plan those separately before asking for a complex sequence.
The camera behaves unexpectedly
Check for conflicting instructions. Decide whether the subject moves, the camera moves, or both; describe each explicitly.
An edit changes too much
Narrow the edit to one object or attribute and identify the features to preserve. Review the output rather than treating preservation wording as a guarantee.
The wording is mostly negatives
Describe the desired appearance directly, such as a sharp subject against a soft background. Do not assume every model handles exclusions the same way.
You want to compare two versions
Keep the source asset and settings unchanged, then alter one instruction. Save the exact prompt beside each output. One result alone does not establish a reliable rule.
Explore more Grok Imagine prompts
For more finished scenes and their original prompts, visit the Grok Imagine prompt library. Use this guide to understand and adapt the wording.