FreeSliders: Training-Free, Modality-Agnostic Concept Sliders for Fine-Grained Diffusion Control in Images, Audio, and Video

TL;DR

FreeSliders introduces a training-free, modality-agnostic method for fine-grained concept control in diffusion models across images, audio, and video, enhancing controllability without additional training.

Contribution

It proposes a novel, training-free approach for concept sliders that works across multiple modalities, extending the evaluation benchmark and addressing scale and traversal issues.

Findings

Enables plug-and-play concept control without training.

Outperforms existing baselines in multi-modal fine-grained control.

Provides new evaluation metrics and tools for controllable generation.

Abstract

Diffusion models have become state-of-the-art generative models for images, audio, and video, yet enabling fine-grained controllable generation, i.e., continuously steering specific concepts without disturbing unrelated content, remains challenging. Concept Sliders (CS) offer a promising direction by discovering semantic directions through textual contrasts, but they require per-concept training and architecture-specific fine-tuning (e.g., LoRA), limiting scalability to new modalities. In this work we introduce FreeSliders, a simple yet effective approach that is fully training-free and modality-agnostic, achieved by partially estimating the CS formula during inference. To support modality-agnostic evaluation, we extend the CS benchmark to include both video and audio, establishing the first suite for fine-grained concept generation control with multiple modalities. We further propose three evaluation properties along with new metrics to improve evaluation quality. Finally, we identify an open problem of scale selection and non-linear traversals and introduce a two-stage procedure that automatically detects saturation points and reparameterizes traversal for perceptually uniform, semantically meaningful edits. Extensive experiments demonstrate that our method enables plug-and-play, training-free concept control across modalities, improves over existing baselines, and establishes new tools for principled controllable generation. An interactive presentation of our benchmark and method is available at: https://azencot-group.github.io/FreeSliders/