Research map / Image generation / Text-to-image diffusion
OmniText: A Training-Free Generalist for Controllable Text-Image Manipulation
Agus Gunawan, Samuel Teodoro, Yun Chen, Soo Ye Kim, Jihyong Oh, Munchurl Kim
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Self-Attention Inversion to TextDiff-2 for text removal (ablation) | TextDiff-2 · SCUT-EnsText all text removal | PSNR: 25.42 -> 30.02 | Improved |
| add Cross-Attention Reassignment on top of SAI (ablation) | OmniText text removal · SCUT-EnsText all text removal | PSNR: 30.02 -> 29.52 | Hurt |
| add Cross-Attention Reassignment on top of SAI (ablation) | OmniText text removal · SCUT-EnsText largest text removal | FID: 16.70 -> 15.33 | Improved |
| add cross-attention content loss L_C via latent optimization (ablation) | TextDiff-2 · ScenePair text editing | word accuracy (ACC, %): ACC 76.41 -> 88.52; MS-SSIM 35.56 -> 29.90 | Mixed |
| add grid trick and self-attention style loss (ablation) | TextDiff-2 · ScenePair text editing | MS-SSIM (x10^-2): 35.56 -> 40.19; ACC 76.41 -> 78.28 | Mixed |
| combine content loss with grid and style loss (ablation) | OmniText · ScenePair text editing | word accuracy (ACC, %): 78.28 -> 78.44 | Improved |
| apply OmniText to replace backbone TextDiff-2 on text editing | TextDiff-2 · ScenePair text editing | MSE (x10^-2): 6.49 -> 4.79 | Improved |
| add grid trick with self-attention modulation to UDiffText | UDiffText · OmniText-Bench style transfer (qualitative) | style transfer quality | Mixed |
| use prompt-tuning for general editing baselines | open-source editing models · ScenePair text editing | word accuracy | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- DreamText — DreamText: High Fidelity Scene Text Synthesis (CVPR 2025) · better on 0 of 1 shared result
- UDiffText — UDiffText: A Unified Framework for High-quality Text Synthesis in Arbitrary Images via Character-aware Diffusion Models (ECCV 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- TextCtrl: Diffusion-based Scene Text Editing with Prior Guidance Control (NeurIPS 2024 · similarity 0.89)
- Large-Scale Text-to-Image Model with Inpainting is a Zero-Shot Subject-Driven Image Generator (CVPR 2025 · similarity 0.86)
- Text Image Inpainting via Global Structure-Guided Diffusion Models (AAAI 2024 · similarity 0.86)
- Prompt Augmentation for Self-supervised Text-guided Image Manipulation (CVPR 2024 · similarity 0.86)
- An Item Is Worth a Prompt: Versatile Image Editing with Disentangled Control (AAAI 2025 · similarity 0.86)
- SteerVTE: Seamless Video Text Editing with Style and Glyph Control (NeurIPS 2026 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
