Training-Free Diffusion Priors for Text-to-Image Generation via Optimization-based Visual Inversion
This work addresses the need for more efficient and accessible text-to-image generation methods, offering a zero-shot alternative that could reduce computational costs, though it is incremental as it builds on existing diffusion models.
The paper tackles the problem of computationally expensive and data-intensive diffusion priors in text-to-image generation by proposing Optimization-based Visual Inversion (OVI), a training-free alternative that uses iterative optimization with novel constraints, achieving quantitative scores comparable to or higher than state-of-the-art data-efficient priors on Kandinsky 2.2.
Diffusion models have established the state-of-the-art in text-to-image generation, but their performance often relies on a diffusion prior network to translate text embeddings into the visual manifold for easier decoding. These priors are computationally expensive and require extensive training on massive datasets. In this work, we challenge the necessity of a trained prior at all by employing Optimization-based Visual Inversion (OVI), a training-free and zero-shot alternative, to replace the need for a prior. OVI initializes a latent visual representation from random pseudo-tokens and iteratively optimizes it to maximize the cosine similarity with the input textual prompt embedding. We further propose two novel constraints, a Mahalanobis-based and a Nearest-Neighbor loss, to regularize the OVI optimization process toward the distribution of realistic images. Our experiments, conducted on Kandinsky 2.2, show that OVI can serve as an alternative to traditional priors. More importantly, our analysis reveals a critical flaw in current evaluation benchmarks like T2I-CompBench++, where simply using the text embedding as a prior achieves surprisingly high scores, despite lower perceptual quality. Our constrained OVI methods improve visual fidelity over this baseline, with the Nearest-Neighbor approach proving particularly effective. It achieves quantitative scores comparable to or higher than the state-of-the-art data-efficient prior, underscoring the potential of optimization-based strategies as viable, training-free alternatives to traditional priors. The code will be publicly available upon acceptance.