SAVIER improves cultural accuracy in Vietnamese text-to-image generation
SAVIER: Source-Anchored Visual Inspection and Evidence-Guided Repair for Vietnamese Cultural Text-to-Image Generation
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Generating images from text about Vietnamese culture can miss important details that identify cultural elements. The researchers created SAVIER, a system that looks at the image it produces and fixes errors related to named Vietnamese cultural entities without changing other requested scene details. SAVIER uses a step-by-step approach with specialized roles to compare the image with the text and reference photos, then makes targeted improvements in one quick pass. Tests show it raises the correctness of these cultural depictions significantly compared to prior methods.
What this means in practice
- •For graphic designers: Create culturally accurate images for Vietnamese-themed projects by automatically refining generated art based on culture-specific cues and references.
- •For advertising teams: Generate culturally relevant visuals in a single pass for campaigns targeting Vietnamese audiences while preserving client scene requests.
Authors
Gia-Huy Thai, Duc-Tien Khong, Trung-Nghia Le
Abstract
We present SAVIER, a training-free, one-pass framework that repairs the depiction of a named Vietnamese cultural entity in a generated image while preserving the scene details requested by the user. The problem is hard because a plausible scene can miss the cues that identify the entity, and prompt-level refinement never inspects the image it produces. SAVIER starts from a Culture-TRIP expanded prompt; a Curator separates source-linked cultural cues from prompt-preservation requirements, a prompt-blind Observer describes the first image, and a Refiner turns the discrepancies into at most three positive instructions before one regeneration with the same seed and a fixed bank of reference photographs. On the 50 bilingual prompts of ViCuP, evaluated with SDXL and FLUX.1-dev against a frozen, human-verified visual contract, SAVIER raises contract fulfillment from 55.6 to 69.6 and from 59.1 to 67.3 over Culture-TRIP and raises CAIRE cultural relevance on FLUX.1-dev. A reference-only ablation that removes the agent repair shows that SAVIER's photographs account for most of the contract gain; the agent roles add country-level relevance on FLUX.1-dev without lowering adherence to the original prompt under the official VQAScore. Project page: https://oxyzgiahuy.github.io/savier/