Generating an Image From 1,000 Words: Enhancing Text-to-Image With Structured Captions

2025-11-10

AI Summary

Current text-to-image models rely on short prompts to generate complex images, resulting in a substantial semantic gap and poor controllability—limiting their applicability in professional settings requiring fine-grained control. To address this, we propose FIBO: (1) the first large-scale, open-source training dataset grounded in fine-grained, structured long descriptions; (2) DimFusion, a lightweight intermediate token fusion mechanism that enables efficient alignment between linguistic representations and visual generation; and (3) Text-as-a-Bottleneck Reconstruction (TaBR), a novel evaluation protocol enabling joint quantitative assessment of long-text control fidelity and image reconstructability. Experiments demonstrate that FIBO achieves state-of-the-art prompt alignment performance among open-source models, significantly improving generation quality, attribute disentanglement, and controllability under long-text guidance.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Text-to-image models have rapidly evolved from casual creative tools to professional-grade systems, achieving unprecedented levels of image quality and realism. Yet, most models are trained to map short prompts into detailed images, creating a gap between sparse textual input and rich visual outputs. This mismatch reduces controllability, as models often fill in missing details arbitrarily, biasing toward average user preferences and limiting precision for professional use. We address this limitation by training the first open-source text-to-image model on long structured captions, where every training sample is annotated with the same set of fine-grained attributes. This design maximizes expressive coverage and enables disentangled control over visual factors. To process long captions efficiently, we propose DimFusion, a fusion mechanism that integrates intermediate tokens from a lightweight LLM without increasing token length. We also introduce the Text-as-a-Bottleneck Reconstruction (TaBR) evaluation protocol. By assessing how well real images can be reconstructed through a captioning-generation loop, TaBR directly measures controllability and expressiveness, even for very long captions where existing evaluation methods fail. Finally, we demonstrate our contributions by training the large-scale model FIBO, achieving state-of-the-art prompt alignment among open-source models. Model weights are publicly available at https://huggingface.co/briaai/FIBO

Problem

• Training text-to-image models with long structured captions for enhanced controllability
• Proposing DimFusion to efficiently process extended textual inputs without increasing tokens
• Introducing TaBR protocol to evaluate image reconstruction accuracy from detailed captions

Innovation

• Training model on long structured captions
• Using DimFusion to integrate lightweight LLM tokens
• Introducing TaBR protocol for controllability evaluation