Video generative models have achieved remarkable progress in synthesizing photorealistic video sequences. However, enabling broader and more creative downstream applications requires fine-grained instance-level video editing, including object insertion, object removal, and texture editing, which has emerged as a prominent yet challenging problem. Existing approaches either propose unified generative frameworks with only coarse semantic control, or design task-specific frameworks for individual editing tasks, limiting their flexibility and applicability across diverse real-world scenarios. To address these limitations, we propose AlbedoEdit, a unified generative video editing framework that jointly supports object insertion, object removal, and texture editing. Our key insight is that the intrinsic albedo map, which is invariant to lighting and contains no specularity, shadowing and inter-reflection effects, provides an effective and user-friendly mechanism for specifying fine-grained appearance edits. Built upon video foundation models, AlbedoEdit is fine-tuned to translate source RGB videos into edited RGB videos, conditioned on a user-edited first-frame albedo. Trained on a new paired synthetic dataset covering all three editing tasks, AlbedoEdit implicitly learns to harmonize edited contents and simulate complex real-world visual effects triggered by editing operations, including specular highlights, soft shadows, and mirror reflections. AlbedoEdit demonstrates superior performance over state-of-the-art video editing approaches, both qualitatively and quantitatively.
Method overview. Given an input RGB video Vs, our video editing framework synthesizes an edited RGB video Vt, conditioned on the input video, a text prompt, and two intrinsic albedo images, the extracted albedo a0 from the first video frame, and the edited albedo, a1. The input video and albedo images are all VAE-encoded to latent space and concatenated along the frame dimension together with the noisy target latent. The text prompt (describing the edited video) is provided through cross-attention, following Wan2.1. The finetuned DiT denoises the latent. Finally, the denoised latent is decoded back to RGB space to produce the edited video.
Free-form edits on real-world videos across all three tasks. Drag the slider to compare Input (left) and Ours (right).
We compare against PISCO on in-the-wild object insertion. Use the left slider to reveal PISCO from the Input, and the right slider to reveal Ours from PISCO.
@misc{zhou2026albedoedit,
title={AlbedoEdit: Unified Instance-Level Video Editing with Albedo Guidance},
author={Xilong Zhou and Bao-Huy Nguyen and Zheng Zeng and Jacob Munkberg and Jon Hasselgren and Thomas Leimkühler and Nima Kalantari and Miloš Hašan and Christian Theobalt},
year={2026},
archivePrefix={arXiv},
primaryClass={cs.GR},
}