Overview

PixVerse Modify is a video editing model for changing existing footage with text instructions, optional masks, and optional reference images.

It is best suited to workflows where the source video already exists and the goal is to replace subjects, add or remove elements, change scene attributes, replace embedded text, or restyle the clip while preserving the original timing and overall structure.

Strengths

Mask-Aware Subject Swaps

PixVerse Modify supports targeted subject replacement using masks and reference images. This makes it useful for swapping one or several people or objects in a clip while keeping the underlying motion and scene timing intact.

Add and Remove Edits

The model can insert new elements into a video or remove existing elements and fill the background after removal. This is useful for product placements, cleanup, prop changes, and selective scene edits.

Free-Form Scene Modification

PixVerse Modify supports broader prompt-driven edits that change scene attributes such as lighting, weather, season, and other visual conditions. It works well when the goal is to transform the mood or setting without rebuilding the clip from scratch.

In-Video Text Replacement

The model can detect and replace embedded text inside the video. This makes it useful for versioning marketing clips, changing signage, or localizing short-form content.

Full-Clip Restyling

PixVerse Modify also supports style transfer across the video, including changing the clip into different visual styles such as comic, ink painting, 2D, or 3D-inspired looks.

Capabilities

Video-to-Video Editing

PixVerse Modify accepts an existing video as the source and generates a modified version of that clip. You can provide a previously generated PixVerse video ID or upload an external video file.

Reference-Guided Editing

Reference images can be used for swap and add workflows, letting the model insert or replace subjects based on uploaded visual targets.

Mask-Based Control

The model supports mask-driven editing for more targeted modifications. Masks can come from PixVerse's own mask workflow or from a user-provided mask image.

Keyframe-Aware Editing

Editing can be anchored to a specific keyframe so that masks and edit targets line up with the frame selected for modification.

Input and Output

• AIR ID: pixverse:modify@1
• Input: source video, text prompt, and optional masks, keyframe selection, and reference images
• Output: edited video clip
• Input limits: up to 1920p, 100 MB, and 30 seconds for uploaded source videos

Best Fit

• Subject replacement and character swaps
• Object cleanup and selective removals
• Adding props or branded elements into existing footage
• Restyling an entire clip into a new visual look
• Changing text, weather, lighting, or scene mood in an existing video