TimeColor: Flexible Reference Colorization via Temporal Concatenation

TL;DR

TimeColor is a video colorization model that uses multiple heterogeneous references with temporal concatenation and attention mechanisms to improve color fidelity, consistency, and stability.

Contribution

It introduces a novel approach to multi-reference video colorization using explicit region assignment, temporal concatenation, and specialized attention to enhance performance.

Findings

Improves color fidelity over prior methods.

Enhances identity consistency and temporal stability.

Supports heterogeneous, variable-count references.

Abstract

Most colorization models condition only on a single reference, typically the first frame of the scene. However, this approach ignores other sources of conditional data, such as character sheets, background images, or arbitrary colorized frames. We propose TimeColor, a sketch-based video colorization model that supports heterogeneous, variable-count references with the use of explicit per-reference region assignment. TimeColor encodes references as additional latent frames which are concatenated temporally, permitting them to be processed concurrently in each diffusion step while keeping the model's parameter count fixed. TimeColor also uses spatiotemporal correspondence-masked attention to enforce subject -- reference binding in addition to modality-disjoint RoPE indexing. These mechanisms mitigate shortcutting and cross-identity palette leakage. Experiments on Sakuga-42M under both single- and multi-reference protocols show that TimeColor improves color fidelity, identity consistency, and temporal stability over prior baselines. Our project page is available at https://bconstantine.github.io/TimeColor/.