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

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/. Diffusion models have emerged as state-of-the-art generative models, capable of producing realistic and diverse outputs across images, audio, and video (Rombach et al., 2022; Ho et al., 2022; Shi et al., 2023). Beyond generating high-quality samples, a central task is controllable generation, the ability to steer the generative process along user-specified signals (Liu et al., 2023; Ho et al., 2022). In particular, text-to-x, where x is a certain modality, has emerged as a powerful control signal for generative models, offering an intuitive human interface and enabling semantically aligned control (Zhang et al., 2023a;b). This text-guided capability plays a central role in creative applications, allowing users to produce high-quality content without requiring technical knowledge or professional design skills.