Write AI video prompts for text to video, image animation and first/last-frame transitions. Copy editable templates and open them in the matching generation mode.
Start creatingCompare the starting image with a playable reference video, then copy its source prompt. Each card identifies where the sample came from.
Input image
Sample video
Upload a photo of two people in an open setting, with enough room for their arms and movement.
A couple playing hand in hand in the sea of flowers.
Try a variation: Try a slower walk before adding a camera move.
Starting setup: Wan 3.0 Video · 9:16. Check resolution, duration and the credit quote in Generate.
Use this promptInput image
Sample video
Choose a clearly visible small animal with a readable floor or path around it.
the handheld camera tracks the mouse as it scurries away.
Try a variation: Replace handheld tracking with a steady low-angle follow shot.
Starting setup: Wan 3.0 Video · 16:9. Check resolution, duration and the credit quote in Generate.
Use this promptDescribe the subject, setting and action when you have no starting image. Add framing and a camera move so the model has a clear shot to build.
Text to video template →Use the photo or illustration to establish the look. Direct what changes over time: a blink, rising steam, moving fabric or a gentle camera move.
Image to video prompts →Provide a first and last image when both compositions matter. Describe the movement between them and use consistent subjects and lighting.
First/last-frame template →Start with one shot and one main action. For a text-based scene, name the subject and setting before directing movement. With a starting image, spend more of the prompt on motion because the image already supplies the composition. Choose a camera instruction that supports the action, then add only the atmosphere that matters.
A useful working brief is: subject and setting + action + camera behavior + pace + light. Begin with a short version, review the result, then change one instruction. Stacking a pan, an orbit, a zoom and several actions makes it difficult to tell which direction helped.
For additional background, see Runway’s image-to-video prompting guide. Its advice is model-specific; use the structure as a starting point and evaluate results with the model you select here.
These original templates are editable briefs, not verified output examples. Links open Wan 3.0 Video with the matching input mode and frame. Upload any required images, then choose duration and resolution and review the credit quote before generating.
Start from text when you want the model to invent the scene. No uploaded image is needed.
A wide street-level shot of an empty coastal lane at sunrise. A bicycle leans against a pale blue wall. The camera tracks slowly forward as a linen curtain moves in an open window. Warm side light, muted colors, natural live-action texture, one continuous shot.
Try a variation: Change sunrise to overcast morning while keeping the same scene and camera move.
Starting setup: Wan 3.0 Video · 16:9. Check resolution, duration and the credit quote in Generate.
Use this promptUpload a close view as the first frame and a wider view of the same room as the last frame. Match lighting and furniture across both images.
The camera moves slowly backward from the opening close-up to reveal the full room shown in the final frame. Daylight stays consistent, the furnishings retain their positions and the movement follows one smooth path. The shot settles into the final composition.
Try a variation: Reduce the difference in framing if the model invents furniture between the two views.
Starting setup: Wan 3.0 Video · 16:9. Check resolution, duration and the credit quote in Generate.
Use this promptA portrait needs different direction from a landscape or product image. Keep the first attempt restrained so you can judge whether the source image, action and camera work together.
Text-to-video prompts establish a scene as well as its motion. Image-to-video prompts accompany an uploaded frame, so they can focus on what the subject and camera should do over time.
Yes. The first/last-frame template opens Wan 3.0 Video in that mode. Upload both frames in Generate. Model support varies; changing models can change the available input modes.
No. Results depend on the selected model, source images and generation settings. Review the output for distortion and continuity. Returning to a similar pose can support editing but does not guarantee a seamless loop.
Copying and adapting the text is free. Creating a video requires an account and sufficient credits. Generate shows the quote for your selected model, duration, resolution and other options before submission.
