Infinite-Story: A Training-Free Consistent Text-to-Image Generation

2025-11-17

AI Summary

To address identity and style inconsistency in text-to-image generation across multiple prompts, this paper proposes a training-free, inference-time consistency control framework. The method employs a scale-autoregressive architecture and introduces a novel identity prompt replacement mechanism to mitigate contextual bias in the text encoder; it further incorporates unified attention guidance and adaptive style injection modules to ensure cross-prompt identity preservation and style stability. All operations are performed solely during inference, requiring no model fine-tuning. Experiments demonstrate that the approach significantly improves image-text consistency while maintaining prompt fidelity, achieving state-of-the-art performance. It processes images at 1.72 seconds per sample—over six times faster than the current fastest method.

Technology Category

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

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

We present Infinite-Story, a training-free framework for consistent text-to-image (T2I) generation tailored for multi-prompt storytelling scenarios. Built upon a scale-wise autoregressive model, our method addresses two key challenges in consistent T2I generation: identity inconsistency and style inconsistency. To overcome these issues, we introduce three complementary techniques: Identity Prompt Replacement, which mitigates context bias in text encoders to align identity attributes across prompts; and a unified attention guidance mechanism comprising Adaptive Style Injection and Synchronized Guidance Adaptation, which jointly enforce global style and identity appearance consistency while preserving prompt fidelity. Unlike prior diffusion-based approaches that require fine-tuning or suffer from slow inference, Infinite-Story operates entirely at test time, delivering high identity and style consistency across diverse prompts. Extensive experiments demonstrate that our method achieves state-of-the-art generation performance, while offering over 6X faster inference (1.72 seconds per image) than the existing fastest consistent T2I models, highlighting its effectiveness and practicality for real-world visual storytelling.

Problem

• Addresses identity inconsistency in multi-prompt text-to-image storytelling
• Solves style inconsistency across different prompts in visual generation
• Eliminates need for fine-tuning while maintaining prompt fidelity

Innovation

• Training-free framework for consistent text-to-image generation
• Identity Prompt Replacement mitigates context bias in encoders
• Unified attention guidance ensures global style consistency