Reaction shot
A character hears an unexpected sound, freezes, and slowly turns.
Your character is already in the story. Generate the action or reaction shot you are missing.
This workflow replaces a selected section of your source video. References guide appearance but do not guarantee exact identity or continuity.
This real demo features an AI-generated cat and a teapot. It demonstrates general scene replacement and timing, not verified human-character consistency.
You've spent time creating your character and assembling a short video. But a reaction shot, entrance, or action is missing. Work from your existing clip and generate a new scene for the moment that needs attention.
Original video
Original footage → unwanted or missing action → original next scene
Updated video
Original footage → AI-generated character action → original next scene
This is scene replacement, not character compositing over untouched frames.
A character hears an unexpected sound, freezes, and slowly turns.
A detective approaches a doorway and pauses before entering.
A character reads a letter, takes a breath, and looks up thoughtfully.
A cloaked traveler emerges from a misty passage as the camera moves closer.
An explorer lifts and inspects an old compass on a table.
A soldier turns, crouches, and moves cautiously toward cover.
An illustrated character notices a glowing object and reaches toward it.
A traveler opens a heavy door and steps into a courtyard.
References provide visual guidance, but facial structure, clothing, accessories, proportions, or style may still change.
Use a clear front or three-quarter view showing the face and outfit.
Add a side view to show hairstyle, costume, and proportions.
Include important costume details, props, or visual style.
Add one short clip to guide movement or camera behavior.
A detective in a dim hallway hears a sound behind a door, slowly turns, and pauses with concern. Keep the reference hairstyle and jacket; cinematic medium shot, subtle handheld camera.
An armored adventurer enters a foggy courtyard, walks slowly, stops, and looks around. Match the visible armor and cloak as closely as possible; smooth tracking camera.
A woman beside a window reads a folded letter, pauses, then raises her eyes thoughtfully. Soft natural light and a gentle camera push-in.
An illustrated fantasy character notices a glowing object, leans forward in surprise, and cautiously reaches out. Match the reference palette and style as closely as possible.
A futuristic soldier moves along an industrial corridor, hears something, turns, and takes cover. Controlled cinematic camera and realistic movement.
The character picks up a compass from a desk, examines it, and puts it back. Stable medium shot, warm indoor lighting.
Describe who appears, one connected action, the setting, and camera behavior. Identity-matching language guides the model but is not a guarantee.
Fill a missing character action between generated scenes.
Explore a replacement performance without regenerating the whole clip.
Create another short performance for a stylized character.
Guide a new shot with approved character references.
Add a character beat to a cinematic sequence.
Try a focused generative replacement within an existing story.
Use an authorized MP4 or MOV clip, 3–60 seconds and up to 50 MB.
Set Start and Resume, describe one action, and optionally add up to three images or one short motion reference.
Choose duration and quality, then inspect identity, motion, transitions, and audio.
Cost depends on the generated scene duration and selected quality. The workspace shows the credit amount before submission. Each generation is a separate paid job; failed jobs follow the applicable credit refund process.
Yes. Upload a video, choose the interval to replace, describe the action, and generate a new scene. The output resumes the original footage after your Resume point.
No. Footage between Start and Resume is replaced; the tool does not composite a character over untouched frames.
Yes. Add up to three supported images to guide appearance, clothing, hairstyle, proportions, or style.
No. Facial features, clothing, proportions, and small design details can change even with references.
Yes. Generate a character-driven replacement for a selected interval and review whether it fits your story.
Yes. Describe the emotion and response. Expression and performance quality may vary.
You can request these actions. Complex motion and object interaction may contain visual or anatomy errors.
Yes. Provide stylized references and describe the intended style. The output may still differ from the original design.
Yes, with footage and references you are authorized to use. Exact likeness and natural movement are not guaranteed.
This is not a person-compositing tool. It generates replacement footage for the selected interval.
Yes. Add up to three reference images, such as different angles or detail views.
Yes. One optional reference video of up to eight seconds can guide motion or camera behavior.
No. Your source video, scene description, and optional references are enough to start.
You can reuse references in separate jobs, but consistent identity across outputs is not guaranteed.
The current workflow generates scenes from 5 to 13 seconds.
Source videos can be 3–60 seconds and up to 50 MB.
No. Start must be at least 1.625 seconds into the source; the earliest slider point is about 1.7 seconds.
The workflow requires at least one second of original footage after Resume, so it cannot replace through the final frame.
No. Generated duration is selected separately, so total output length may change.
Audio inside the replaced interval is not guaranteed to be preserved. Plan to review and edit it if needed.
Describe framing, movement, lighting, and environment in the prompt. The model may not follow every instruction precisely.
No. This workflow requires a source video. Use a video or character animation generator to start from scratch.
No. One generation replaces one selected interval. Multiple scenes require separate jobs and editing.
Standard is 480p and High is 768p.
Cost depends on generated duration and quality. The workspace displays the credit amount before submission.
Failed jobs follow the service's applicable credit refund process. A completed result needing revisions is not automatically a failed job.
Use only source material and references you have rights or consent to use, and follow the applicable service terms.
No. This is generative scene replacement with visual reference guidance, not identity locking or face swapping.
Choose the part of your existing video that needs another action, guide the AI with character references, and generate a new scene for your story.
Generate character sceneMP4 or MOV · Up to 3 character references · 5–13 second scenes · Review identity, transitions, and audio before delivery.
Creative Toolkit Index
Quick access to precision layer extraction, AI photo editing & asset tools.