Video face swap is the Edit Video tool that puts a portrait face onto the person in a video. It uses your selected model; in No Model mode, you provide a required Reference image. You provide a source video, and it replaces the face throughout while keeping the original motion and scene.
What does Video face swap do?
It swaps the face in your source video for your currently selected portrait model's face. In No Model mode the face comes from the Reference image you select or upload. The rest of the video stays as it is. The tool's hint sums up the source input: "Select or upload the video you want to use for the face swap."
How do I run a Video face swap?
Open Edit Video and pick Video face swap from the tool switcher.
Add the video with Select content or Upload video.
In No Model mode, add the required Reference image with Select content or Upload image. This card is hidden when a portrait model is selected.
Adjust quality in the bottom bar if you want to change it.
Click Generate — the cost is shown next to the button.
Which face gets used?
Your selected portrait model's face, or the required Reference image in No Model mode. The two flows stay separate: model-backed generation does not show or submit the extra image.
What quality and resolution options do I get?
Video face swap shows a quality picker but no resolution picker — resolution selection is hidden for this tool. Best supports a selected clip up to 15 seconds, while Ultra supports up to 30 seconds. If a clip longer than 15 seconds is selected on Ultra, switching to Best is blocked with "Trim the video to 15 seconds or less before switching quality." Use the bottom-bar trimmer to shorten it, then switch again. The per-second cost shown next to Generate updates with the quality and depends on your plan.
Why is Generate disabled?
Generate stays off until your source video has finished uploading. In No Model mode it also waits for the Reference image selection/upload; if that image is missing, the app says "Please select or upload a reference image." With a portrait model selected, no extra image is required.
What shows on the detail page, and can I Reuse it?
The detail page shows the Source video and, when No Model generation supplied one, the Reference image. Reuse restores both saved inputs. Older records generated before Reference image detail data was saved omit it and require you to select that input again.
On mobile
The source-video flow is the same. In No Model mode, the required Reference image card is appended after the source-video card; with a portrait model selected it stays hidden.
Related: Edit Video overview · Edit character · Upscale