Build a scene from multimodal references
Use images to establish the look, video to demonstrate an action, or audio to set the rhythm. A storyboard, creative brief, or public web page can supply the background and direction for the scene.
Wan 3.0 is Alibaba’s multimodal AI video model for longer stories, consistent characters and scenes, and precise video edits. Create clips up to 30 seconds in 1080p on GetReelMaker, with control over references, camera direction, and sound.
Use images to establish the look, video to demonstrate an action, or audio to set the rhythm. A storyboard, creative brief, or public web page can supply the background and direction for the scene.
Give a scene time to establish the setting, develop an action, move between story beats, and reach a conclusion. Choose the length yourself or let Wan 3.0 select it. Any reference footage counts toward the 30-second total.
Reference materials help preserve a character’s appearance, wardrobe, props, and sound cues as the action changes. They can also guide continuity in the layout of the scene and the positions of subjects relative to one another.
Provide the original clip as a video reference. Describe which subject, background, lighting, or style to change and which details to preserve. To extend the clip, describe what should happen after its final shot.
Use the audio setting to generate sound alongside the picture. Describe the voice, regional speech, musical rhythm, or ambient sound and how it should follow the performance. Audio references can guide the mood or pacing.
Specify the lighting, color, framing, and camera movement, down to the timing of a gesture or how an object responds on impact. Direct the performance and describe how materials such as fabric, glass, or metal should look and move.
The GetReelMaker creator connects to the standard Wan 3.0 Video model via KIE. Use public URLs for image, video, audio, document and web page inputs.
Turn a shot idea into a video in three steps.
A low tracking shot follows a white race car drifting around a wet city corner at night. Streetlights reflect off the asphalt as tire smoke trails behind.
Write a prompt in the creator above. For keyframes, media references, a document, or a web page, paste the public URL for each input you want to use.
Set the duration, resolution, aspect ratio and audio. Get a credit estimate before submitting.

Sign in to generate and preview your clip. Adjust one detail at a time if you want to compare takes.
Pick the one object or person the viewer should notice first.
Use a concrete direction such as overhead follow, push-in, or a fixed shot.
Give the scene a light source and a direction, such as late sun from the right.
Use a short duration for a single movement. Allow more time for actions or transitions that need to unfold.
Try two versions with the same subject and a different angle to compare the shots.
Wan 3.0
Wan 3.0 is Alibaba’s multimodal video generation model. KIE’s standard Wan 3.0 Video model can create new clips from text or images, use media references, and edit or extend existing footage.
KIE supports 2–30 seconds and resolutions up to 1080p. A request with reference video must fit the input and output into 30 seconds total.
KIE’s Wan 3.0 Video API can generate sound with the picture and accepts audio references in reference mode. Use the audio switch to choose whether to include sound.
KIE supports editing or continuing footage through video references. Supply the clip and describe the intended change or what should happen next.
KIE supports a first image to anchor the shot and an optional last image to guide the ending. Reference media, files, and web links are separate from the keyframe workflow.
Choose text to video, first and last frames, media references, a document or a public web page. Provide publicly accessible URLs for reference inputs.
Yes. Sign in and have enough credits to submit a generation. Request a credit estimate before submitting.