Generate and edit video

You can generate a video from a text prompt or use images, video, and audio nodes as references. Once the video is ready, select its node to open the Toolbar. You can trim it, replace or remove objects, continue the footage, retake shots from new angles, capture frames, preview, or download it.

Generate a video

1. Create a video node from the Add menu, or ask Agent to create one;
2. Describe the subject, main action, setting, and camera movement;
3. Connect reference nodes or mention them with @;
4. Choose a model, aspect ratio, duration, resolution, and any other settings available for that model;
5. In Ask mode, review the inputs and settings on the confirmation card;
6. Select Generate;
7. Play the finished video from beginning to end.

Here is a prompt you can use:

Describe the subject's movement first, then describe the camera movement. Keeping each clip focused on one main action usually makes the result easier to control.

The available references, duration, audio, and output settings vary by model. Check the options shown in the video node. The Tapies shown on the page are an estimate; final usage is based on actual settlement.

Generate a Seedance 2.5 video at 1080p

1. Create or select a video node, then choose Seedance 2.5. You can also ask Agent to create a video and choose the model on its confirmation card.
2. Open Generate method, then choose Frames, Omni Reference, or Video Edit for the task.
3. Open Resolution and select 1080p.
4. Check the inputs, remaining settings, and updated Tapies estimate, then generate the video.

All of these generation methods support 1080p. New tasks still start at 720p, so select 1080p before each generation that needs the higher resolution. Omni Reference accepts up to 30 images, 10 videos, and 10 audio clips; follow the methods and limits shown in the current node. When generation finishes, the 1080p result appears in the current video node.

Seedance 2.5 requires a paid account and access to the model. It appears locked when it is not available for the current account.

Generate video with MiniMax H3

1. Create or select a video node, then choose MiniMax H3;
2. Open Generate method, choose Frames or Reference, then enter a prompt and add the references you need;
3. Choose 768P or 2K and set a duration from 4 to 15 seconds. New tasks start at 2K and 5 seconds;
4. Check the estimated Tapies shown on the page, then select Generate.

MiniMax H3 generates video with native audio. Reference mode accepts up to nine images, three videos, and three audio clips. Each reference video must be 2–15 seconds, and their combined duration cannot exceed 15 seconds. Audio cannot be used by itself; add at least one image or video. MiniMax H3 requires a paid account and access to the model; it appears locked when unavailable. Play the finished result from beginning to end to check the visuals, audio, and subject consistency.

Draft faster with MiniMax H3 Max

MiniMax H3 Max is the high-speed option in the H3 family. Use it when you want to compare motion, composition, or prompt directions quickly before choosing a take to refine.

1. Create or select a video node and choose MiniMax H3 Max. You can also switch to it in an Agent confirmation card.
2. Choose text-to-video, or add one image as the first frame.
3. Choose 480P or 768P and set a duration from 5 to 15 seconds.
4. Describe the subject’s action, camera movement, and intended sound. Check the estimated Tapies, then generate.

MiniMax H3 Max generates video with native audio. It currently supports text-to-video and image-to-video from one starting frame, but not first-and-last-frame or multimodal reference generation. Completion time still varies with the queue and input. The options shown in the current node are the source of truth.

Generate video with Wan 3.0

1. Create or select a video node, then choose Wan 3.0.
2. Choose text generation, first-frame generation, or reference-image generation. Reference mode accepts 1–10 images.
3. Choose 480P, 720P, or 1080P, then set a duration from 2–30 seconds or choose Auto.
4. Keep or turn off audio, check the estimated Tapies, then generate.

Text and reference-image generation support Adaptive, 16:9, 9:16, 1:1, 4:3, and 3:4. First-frame generation uses the input image’s aspect ratio. Wan 3.0 generates audio by default. The model and available settings appear according to your account.

Generate video with FLUX 3

1. Create or select a video node, then choose FLUX 3.
2. Enter a text prompt or add one image as the first frame.
3. Choose 720p or 1080p and set a duration from 5–20 seconds.
4. Keep or turn off audio, check the estimated Tapies, then generate.

Text generation supports 21:9, 2:1, 16:9, 4:3, 1:1, 3:4, and 9:16. First-frame generation uses the input image’s aspect ratio. FLUX 3 generates audio by default. The model and available settings appear according to your account.

Open the video toolbar

After a video has been generated or uploaded:

1. Select the video node;
2. The available tools appear in the node Toolbar;
3. Hover over an icon to see its name;
4. Select a tool and follow the controls in its panel.

Some tools may be hidden while a video is generating, when the canvas is read-only, or when the feature is unavailable for the current account.

Core toolbar tools

Tool	When to use it	What happens
Trim	Remove the beginning or end, or keep one section of a video	Opens the trim panel. After you set the in and out points and confirm, Creative OS creates a new trimmed video node and keeps the original
Replace	Replace an object in the video using a reference image	Opens the replacement editor. After you select the object and a reference image, Creative OS creates a connected result node and keeps the source video
Remove	Remove a person or object from the video	Opens the removal editor. After you select the target, Creative OS creates a connected result node and keeps the source video
Continue video · Beta	Add footage before the beginning or continue the ending	Opens the continuation controls. Creative OS creates a connected result node before or after the source video
Video Retake (Beta)	Regenerate shots with different views, shot scales, or camera movement	Opens the shot timeline and camera controls. Creative OS creates a connected result node and keeps the source video
Capture frame	Reuse one frame as an image or as a reference for the next shot	Captures the current, first, or last frame and creates an image node
Save to library	Keep the video for future work	Saves the video to your personal or team asset library
Download	Save the current video to your computer	Downloads the video file
Full-screen preview	Check motion, visual details, or audio	Opens a larger preview without changing the node

These are the core tools used most often. The buttons shown can vary with the video state, account access, and product version.

Trim

To remove the beginning or end of a video, or keep one section:

1. Select the video node and choose Trim in the Toolbar;
2. Drag the in and out points on the timeline;
3. Play the selected range to check both cut points;
4. Select Confirm Trim.

Creative OS creates the trimmed result beside the source video. The source video stays on the canvas.

For a video that already contains several shots, use Smart Clip in the trim panel. It analyzes changes in the footage, detects clear scene cuts, and turns the detected shots into separate video nodes.

Smart Clip is useful for splitting an edited ad, short film, or storyboard sample without finding every cut point manually. Play each result after the analysis finishes and check the cuts. If the video has no clear visual transitions, Smart Clip may not find usable cut points.

Replace an object

1. Select the video node and choose Replace in the Toolbar;
2. Play or scrub to a frame where the object is clear, select Add Overlay, drag a box around the object, and confirm the selection;
3. Choose a replacement image with From Canvas or From Local;
4. Check the estimated Tapies, then select Generate.

Creative OS uses the selected object and reference image to generate a new, connected video node beside the source. The source video stays on the canvas.

Remove an object

1. Select the video node and choose Remove in the Toolbar;
2. Play or scrub to a frame where the target is clear, select Add Overlay, drag a box around it, and confirm the selection;
3. Check the estimated Tapies, then select Generate.

Creative OS creates a new, connected video node with the selected target removed. The source video stays on the canvas.

Continue a video

1. Select the video node and choose Continue video · Beta in the Toolbar;
2. Choose Add prologue or Continue story;
3. Choose the aspect ratio, an added duration from 4 to 30 seconds, and how the footage should continue. You can describe what happens before or next and use @ to add reference assets;
4. Check the estimated Tapies, then select Confirm and generate.

Creative OS creates a connected video node before or after the source, based on the direction you selected. The source video stays on the canvas.

Video Retake (Beta)

1. Select the video node and choose Video Retake in the Toolbar to open the retake panel;

   

2. On the Shot timeline, select the segment you want to change. If different time ranges need different treatments, move the playhead to each cut point and choose Split cut. You can configure every segment separately;

   

3. Select a segment, then choose its camera treatment:

   • Keep original: use this when the segment does not need a change. Generation is instructed to preserve the source footage without changing the view or shot scale;
   • Fixed camera: retake the full segment from one stationary view. Drag the sphere on the left to place the camera, or choose a preset on the right—Front, Front side, Rear side, Back, Overhead, or Low Angle. Then adjust Shot scale from a close-up to a far shot;
   • Switch views: set a Start view and an End view so the segment cuts directly from one angle to the other, without a continuous camera move. Use it for dialogue coverage, reaction shots, or an emphasized moment;
   • Smooth movement: set a Start view and an End view so the camera transitions continuously between them. Use it to suggest an orbit, push-in, pull-back, or a move from a side view to the front.

   For Fixed camera, use the sphere or view presets to set one camera. For Switch views and Smooth movement, select Start view or End view, then configure each endpoint with the sphere, view presets, and Shot scale. You can also add shot notes to tell the AI what the segment should emphasize. Segments shorter than 0.5 seconds support only Keep original or Fixed camera. If the start and end views are the same or nearly the same, the segment is treated as a fixed camera.

   

4. Check the resolution, audio setting, and estimated Tapies, then select Generate video.

Creative OS uses the camera plan for each segment to guide AI generation and creates a new, connected video node beside the source. The source video stays on the canvas. The sphere, presets, and shot notes express your camera intent; they are not a frame-accurate 3D camera path. Play the result to check the view and camera movement.

Capture a frame

Frame capture turns one moment in a video into a new image node. Select the video, choose Capture frame in the Toolbar, then select:

• Capture current frame: play or scrub to the moment you need, then capture it;
• Capture first frame: use the first frame of the video;
• Capture last frame: use the final frame of the video.

The captured image appears as a new node on the canvas. You can use it to generate the next shot, build a storyboard, or reference it in another image or video.

Next: Generate and edit audio