TokenDial: Continuous Attribute Control in Text-to-Video via Spatiotemporal Token Offsets

TL;DR

TokenDial introduces a method for continuous, attribute-specific control in text-to-video models by learning semantic offsets in token space, enabling predictable and high-quality edits without retraining.

Contribution

It proposes a novel approach to attribute control in text-to-video generation through learned token offsets, enhancing controllability without retraining the entire model.

Findings

Achieves stronger controllability than existing methods.

Produces higher-quality, predictable edits across diverse attributes.

Validated by extensive quantitative and human evaluations.

Abstract

We present TokenDial, a framework for continuous, slider-style attribute control in pretrained text-to-video generation models. While modern generators produce strong holistic videos, they offer limited control over how much an attribute changes (e.g., effect intensity or motion magnitude) without drifting identity, background, or temporal coherence. TokenDial is built on the observation: additive offsets in the intermediate spatiotemporal visual patch-token space form a semantic control direction, where adjusting the offset magnitude yields coherent, predictable edits for both appearance and motion dynamics. We learn attribute-specific token offsets without retraining the backbone, using pretrained understanding signals: semantic direction matching for appearance and motion-magnitude scaling for motion. We demonstrate TokenDial's effectiveness on diverse attributes and prompts, achieving stronger controllability and higher-quality edits than state-of-the-art baselines, supported by extensive quantitative evaluation and human studies.