Pith. sign in

REVIEW 3 major objections 6 minor 28 references

HDGlyph: A Hierarchical Disentangled Glyph-Based Framework for Long-Tail Text Rendering in Diffusion Models

T0 review · 3 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read A diffusion framework that conditions on glyph edge maps renders unseen scripts and tiny text while preserving background quality.

desk verdict Solid engineering with a coherent stack of tricks, but the multilingual benchmark confounds unseen scripts with smaller fonts, which undermines the paper's headline long-tail claim. read the letter →

arxiv 2505.06543 v1 pith:E2Q6H4VV submitted 2025-05-10 cs.CV

classification cs.CV
keywords visualtextrenderingdiffusionmodelslong-tailglyphconditioningmultilingualgenerationsmallclassifier-freeguidancelatentdisentanglement
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper sets out to show that diffusion models fail at rare and tiny text because they never separate text from background, and that forcing that separation at three levels—pixel, noise, and latent—lets a single model handle common text, scripts it never saw in training, and very small fonts. The proposed framework, HDGlyph, conditions generation on the Canny edge map of the target glyph string through a Multi-Linguistic GlyphNet, trained with a glyph-aware perceptual loss that aligns predicted and ground-truth text-region edges instead of relying on OCR feedback. At inference it adds a noise-level guidance term that amplifies the glyph signal against a glyph-free negative condition, and a two-stage latent scheme that paints the background first, then refines the text, with a patch-based upscaling pass for small text. The reported payoff on a public text-rendering benchmark is 5.08% higher English accuracy and 11.7% higher Chinese accuracy than prior open-source systems, with much smaller accuracy drops on Japanese, Russian, and small-text content.

What carries the argument

The load-bearing mechanism is the Multi-Linguistic GlyphNet, a ControlNet-style branch initialized from a Canny-conditioned ControlNet and augmented with per-language expert LoRAs, paired with the Glyph-Aware Perceptual Loss $\mathcal{L}_{\text{Glyph}}$, which compares Canny edge maps of predicted and ground-truth text regions with a noise-aware timestep weighting. Together they supply a script-agnostic structural prior: the loss lives entirely on edge geometry, never on OCR recognition, which is what lets unseen characters transfer to new scripts. At inference, Noise-Disentangled Classifier-Free Guidance (ND-CFG) reweights the guidance equation with a glyph-amplifying coefficient $\omega_{ndg}$ and a glyph-free negative image condition, while Latent-Disentangled Two-Stage Rendering (LD-TSR) re-injects background latents from a first unconditional pass and applies a cosine-scheduled guidance boost that sharpens glyph details late in denoising; for small text, the pipeline denoises overlapping latent patches at up to 2048×2048 resolution, using patch-based denoising, to recover stroke-level clarity.

What would settle it

Render a connected-stroke script such as Arabic or Devanagari at small sizes and measure per-character accuracy against the English baseline: if accuracy collapses toward the levels of the prior baselines, the edge prior is not complete for those stroke structures. A more direct check is to perturb the Canny thresholds used to build the glyph conditions and measure whether character accuracy tracks edge fidelity; if it does, the framework's ceiling is set by the edge extractor rather than the diffusion model.

Watch

Extended reading notes

Core claim

The paper's central claim is that long-tail text failure in diffusion models is a disentanglement problem, not a data or encoder problem. HDGlyph establishes that the Canny edge map of the target glyph string is a script-agnostic structural prior: because edges encode stroke geometry without requiring the model to have memorized the character, a network trained only on English and Chinese glyphs can render Japanese, Russian, and other scripts it never saw, legibly, from their edge maps alone. The framework separates text from non-text at the pixel level with an edge-supervised glyph branch and per-language expert adapters, at the noise level with an extra guidance weight that amplifies the glyph condition against a glyph-free negative image condition, and at the latent level with a two-stage render that first paints the background and then refines the text, plus an upscaled patch-denoising pass for small text. The experiments assert that this hierarchy is what lets one model deliver both high text accuracy and high image quality, where prior systems trade one for the other.

Load-bearing premise

The load-bearing premise is that the Canny edge map of the target text is a complete and transferable structural prior, so a model trained only on English and Chinese glyphs can render unseen scripts and very small fonts from those edge maps alone; if edge extraction loses stroke-level detail for dense or tiny glyphs, the unseen-character and small-text claims weaken.

Editorial extensions

If this is right

  • Text accuracy no longer trades against background quality: the two-stage latent scheme recovers most of the image-quality loss that aggressive glyph guidance costs.
  • Unseen characters become renderable without retraining: the edge prior transfers across scripts, so the English- and Chinese-trained model produces legible Japanese and Russian with far smaller accuracy drops than the prior baselines.
  • Small text is addressed by upscaled patchwise denoising, cutting the measured accuracy drop on the small-text benchmark to 8.07%, versus 13.43% and 10.51% for the closest open-source systems.
  • The design is portable across diffusion architectures: the paper implements it on both a U-Net backbone and a DiT backbone and reports consistent gains on both.
  • Supervising glyphs through edge alignment removes the dependence on OCR models during training, eliminating a source of bias that recognition-based losses introduce.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A testable consequence the paper does not run: if edge maps are script-agnostic, a model trained on English alone should already render Chinese and Japanese legibly; measuring that zero-shot transfer would isolate how much of the gain comes from the edge prior rather than from the English-plus-Chinese training mixture.
  • The appendix records that the small-text patch mechanism relies on latent priors unique to the U-Net backbone and underperforms on the DiT variant, so making the latent disentanglement architecture-neutral is the natural next step.
  • Because the glyph condition is just an edge map, any renderable font becomes a possible conditioning input, suggesting a route to arbitrary scripts and stylized fonts that bypasses the need for paired training data in those scripts.
  • The dependence on thresholded edge maps implies a ceiling: cursive or connected-stroke scripts, whose Canny edges fragment at small sizes, may need a stroke-preserving preprocessing step before the same pipeline can claim full multilingual coverage.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper proposes HDGlyph, a framework for visual text rendering in diffusion models. It hierarchically disentangles text generation from background synthesis at three levels: at training time, a Multi-Linguistic GlyphNet and a Glyph-Aware Perceptual Loss are used to preserve glyph structures; at inference time, Noise-Disentangled Classifier-Free Guidance amplifies glyph-specific signals, and Latent-Disentangled Two-Stage Rendering with a small-text enhancement module refines backgrounds and small glyphs. The authors evaluate HDGlyph on the AnyText-Benchmark, a self-constructed Multilingual-Benchmark, and the Complex-Benchmark, reporting gains of 5.08% in English and 11.7% in Chinese text accuracy, as well as smaller accuracy drops than AnyText and AnyText2 on multilingual and small-text scenarios. The paper also provides ablations showing the contribution of each proposed component.

Significance. If the results hold, HDGlyph is a practical engineering contribution: it improves text accuracy over existing open-source baselines on an external benchmark, and the ablation study gives internal evidence that each proposed component contributes. The use of Canny glyph edges as a transferable structural prior and per-language LoRA experts is a reasonable design. However, the strongest claim—robust generalization to unseen scripts—rests on a self-constructed benchmark with a font-size confound, and the main comparison table mixes results from inconsistent evaluation pipelines. The AnyText-Benchmark results and the internal ablations are credible, but the manuscript needs revision before the long-tail claims are fully supported.

major comments (3)
  1. [Section 4.2.2, Appendix A.1, Table 2] The Multilingual-Benchmark construction conflates script novelty with font size. The paper claims that reusing the English layout for translated text 'eliminates the influence of layout variations and font size differences,' but Appendix A.1 concedes 'we assume that the font sizes are roughly the same.' For a fixed bounding box, a translated string of different length renders at a different font size; no normalization is described. Because Table 2 reports large accuracy drops for Japanese (82.00 to 59.70) and Russian (82.00 to 62.94), these drops could be caused by smaller rendered glyphs rather than by unseen-script difficulty, which is the paper's core long-tail claim. Please measure and report per-sample font sizes (or per-character pixel heights), and rerun with size-normalized conditions—for example, matching rendered text pixel height across languages or reporting results stratified by font size.
  2. [Table 1, Section 4.3.1, Appendix B] The baseline comparison is not on a consistent evaluation pipeline. The authors adopt selected results from [15,20] (marked with †) while other rows are presumably run locally, and the resulting CLIPScore values for Glyph-ByT5 (48.02) and GlyphDraw2 (47.96) are on a completely different scale from the locally run models (roughly 80–90). The paper itself notes that different CLIPScore models produce divergent values, but it still reports these rows in the same table. This undermines the claim that HDGlyph 'maintains high image quality' relative to those methods. Please rerun all baselines under the same protocol, or clearly separate the tables and restrict claims to methods evaluated consistently.
  3. [Section 3.1.1] The 'unseen character' claim rests on the assumption that a Canny edge map of the target glyph is a complete and transferable structural prior for scripts not seen during training. The paper provides no analysis of how Canny edge quality degrades for dense scripts (e.g., Japanese kanji, Russian Cyrillic) or for small text, and the Multilingual-Benchmark cannot resolve this because of the size confound. Please add an edge-fidelity analysis (e.g., character-level IoU between the rendered and ground-truth edge maps across languages and font sizes) or an ablation that holds font size fixed while varying script novelty.
minor comments (6)
  1. [Abstract, Section 3.1.2, Appendix C] The phrase 'Glyph-A ware' should be 'Glyph-Aware,' and the caption of Figure 2 in Appendix C says 'HMDGlyph' instead of 'HDGlyph.'
  2. [Appendix A, Appendix B] Appendix A and B contain unresolved citation placeholders such as '[? ]' and 'work [? ];' these need to be filled before publication.
  3. [Section 4.2.2] The Multilingual-Benchmark is constructed from 100 randomly selected prompts, but no seed or exact list is provided; please supply the list or seed to make the benchmark reproducible.
  4. [Appendix B, Section 4.2.4] Appendix B states 'All values are expressed as percentages,' but it also reports ViT-L/14 CLIPScore values of 0.6912 and 0.6901 for AnyText and AnyText2; please specify the scaling and state which CLIPScore version applies to each table.
  5. [Equation (7), Section 4.1] The explanation that a larger 𝛼2 value 'requires a faster rate' is vague; please specify how 𝛼1 and 𝛼2 change the shapes of the cosine schedules in Eqs. (6) and (7).
  6. [Table 4, Table 3] Table 4's 'ControlNet' baseline should be clarified: it is presumably ControlNet-Canny without expert LoRA, ND-CFG, or LD-TSR, but this is not stated, and the relationship to 'SDXL+Canny' in Table 3 is unclear.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: HDGlyph is an empirical engineering contribution benchmarked on external and held-out data, with no derivation reducing to its own inputs.

full rationale

HDGlyph does not derive a prediction from first principles; it proposes a conditional generation architecture and evaluates it empirically. The glyph image c_I is the task input, not a fitted parameter, and the Glyph-Aware Perceptual Loss supervises predicted versus ground-truth Canny edge maps, which is direct training supervision rather than a disguised form of the evaluation metric. The main accuracy gains (5.08% English, 11.7% Chinese) come from the externally constructed AnyText-Benchmark, and the Multilingual- and Complex-Benchmarks, although self-constructed, are held-out with respect to training and are not used to fit the reported hyperparameters. The hyperparameters (omega_ndg=5.0, alpha1/alpha2=3.0/4.0, A=3.0) are tuned implementation choices, and the paper does not rename tuned values as predictions. The only self-citation ([5], CustomContrast) is a passing related-work reference and is not load-bearing; no uniqueness theorem is imported from the authors' prior work, and no ansatz is smuggled in through a citation. The Multilingual-Benchmark does contain a validity concern: reusing English layouts for translated strings can change rendered font size, and Appendix A.1 merely assumes 'font sizes are roughly the same' while Appendix A.2 shows font size strongly affects accuracy, so the unseen-character results may be confounded with small-text difficulty. That is a benchmark-validity issue, not a circularity: no equation reduces to itself, no fitted quantity is renamed a prediction, and no load-bearing claim rests on a self-citation. Appendix B also discloses a DiT-specific limitation for small-text rendering, which narrows scope but does not create circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim rests on a standard diffusion recipe plus several hand-tuned curves (omega_ndg, A, alpha1, alpha2) and domain assumptions about glyph edges and OCR accuracy. No new physical or conceptual entities are introduced. The free parameters are not derived from theory and lack sensitivity analysis.

free parameters (6)
  • omega_ndg (glyph guidance weight) = 5.0
    Hand-set in Section 4.1; amplifies the glyph-conditioned noise prediction in Eq. (5). No sensitivity analysis is reported.
  • A (dynamic ND-CFG schedule magnitude) = 3.0
    Eq. (7); controls how much additional glyph guidance is applied as denoising progresses.
  • alpha1 (background blend curve exponent) = 3.0 (U-Net), 4.0 (DiT)
    Eq. (6); shapes the cosine mixture c1 between background latents and current latents in LD-TSR.
  • alpha2 (guidance schedule exponent) = 3.0 (U-Net), 4.0 (DiT)
    Eq. (7); shapes the rate at which omega_ndg increases with denoising step.
  • omega_cfg (classifier-free guidance scale) = 7.5
    Set in Appendix B; standard value inherited from prior work, not a novel contribution.
  • phi(t) loss weighting schedule = alpha_bar_t (DDPM) or (1 - sigma_t)^2 (SD3)
    Eq. (2); chosen from prior work based on empirical improvement, not fitted to the target data.
assumptions (5)
  • standard math Diffusion training and classifier-free guidance formulas (Ho et al., Ho and Salimans) are valid.
    Used in Eqs. (1), (3), and (5) without modification.
  • domain assumption Canny edges of a rendered glyph image are a sufficient structural prior for generating text in any script, including scripts unseen in training.
    Core to Multi-Linguistic GlyphNet (Sec. 3.1.1); the claim of generalization to unseen characters rests on this.
  • domain assumption AnyText-3M layout annotations (text bounding boxes) are accurate enough to extract text regions for the glyph-aware loss.
    The Glyph-Aware Perceptual Loss (Eq. (2)) crops predicted and ground-truth images using these annotations.
  • domain assumption EasyOCR gives an unbiased and sufficiently accurate measure of text accuracy across English, Chinese, Japanese, and Russian for the constructed benchmark.
    Appendix B states EasyOCR is used as the evaluation OCR; if it is biased or noisy for certain scripts, reported ACC/NED values are not comparable across languages.
  • ad hoc to paper The cosine-based blend schedules in Eqs. (6)-(7) (c1 and omega_ndg tilde) improve background fidelity and small-text accuracy without a principled justification.
    Introduced for LD-TSR and ND-CFG; no derivation beyond empirical observation. The exponents alpha1 and alpha2 are tuned by hand.

how reviews work

0 comments
Cite this review

Pith. "Pith review of HDGlyph: A Hierarchical Disentangled Glyph-Based Framework for Long-Tail Text Rendering in Diffusion Models." pith.science (2026). https://pith.science/paper/E2Q6H4VV

@misc{pith2026250506543,
  author       = {Pith},
  title        = {Pith review of: HDGlyph: A Hierarchical Disentangled Glyph-Based Framework for Long-Tail Text Rendering in Diffusion Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/E2Q6H4VV}},
  note         = {Machine review of arXiv:2505.06543}
}
read the original abstract

Visual text rendering, which aims to accurately integrate specified textual content within generated images, is critical for various applications such as commercial design. Despite recent advances, current methods struggle with long-tail text cases, particularly when handling unseen or small-sized text. In this work, we propose a novel Hierarchical Disentangled Glyph-Based framework (HDGlyph) that hierarchically decouples text generation from non-text visual synthesis, enabling joint optimization of both common and long-tail text rendering. At the training stage, HDGlyph disentangles pixel-level representations via the Multi-Linguistic GlyphNet and the Glyph-Aware Perceptual Loss, ensuring robust rendering even for unseen characters. At inference time, HDGlyph applies Noise-Disentangled Classifier-Free Guidance and Latent-Disentangled Two-Stage Rendering (LD-TSR) scheme, which refines both background and small-sized text. Extensive evaluations show our model consistently outperforms others, with 5.08% and 11.7% accuracy gains in English and Chinese text rendering while maintaining high image quality. It also excels in long-tail scenarios with strong accuracy and visual performance.

Figures

Figures reproduced from arXiv: 2505.06543 by the authors.

Figure 1
Figure 1. Illustration of our motivation. The curve chart [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Glyph-aware Training Pipeline of HDGlyph.Blue [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Our HDGlyph framework inference pipeline comprises Multi-Linguistic GlyphNet, along with (a) the Noise [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison of HDGlyph with state-of-the-art models in long-tail text rendering of multilingual and [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 1
Figure 1. Figure 1: Using ControlNet-Canny only on the English test set of Multilingual-Benchmark (selecting the first text and its [PITH_FULL_IMAGE:figures/full_fig_p011_1.png]
Figure 2
Figure 2. Figure 2: More Qualitative Results of HMDGlyph [PITH_FULL_IMAGE:figures/full_fig_p012_2.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 13 canonical work pages

  1. [1]

    Omer Bar-Tal, Lior Yariv, Yaron Lipman, and Tali Dekel. 2023. MultiDiffusion: Fusing Diffusion Paths for Controlled Image Generation. In International Con- ference on Machine Learning, ICML 2023, 23-29 July 2023, Honolulu, Hawaii, USA (Proceedings of Machine Learning Research, Vol. 202), Andreas Krause, Emma Brun- skill, Kyunghyun Cho, Barbara Engelhardt,...

  2. [2]

    Haoxing Chen, Zhuoer Xu, Zhangxuan Gu, Jun Lan, Xing Zheng, Yaohui Li, Changhua Meng, Huijia Zhu, and Weiqiang Wang. 2023. DiffUTE: Uni- versal Text Editing Diffusion Model. In Advances in Neural Information Pro- cessing Systems 36: Annual Conference on Neural Information Processing Sys- tems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 202...

  3. [4]

    Jingye Chen, Yupan Huang, Tengchao Lv, Lei Cui, Qifeng Chen, and Furu Wei

  4. [5]

    Nan Chen, Mengqi Huang, Zhuowei Chen, Yang Zheng, Lei Zhang, and Zhendong Mao. 2024. CustomContrast: A Multilevel Contrastive Perspective For Subject- Driven Text-to-Image Customization. arXiv preprint arXiv:2409.05606 (2024)

  5. [6]

    Hospedales, Yi-Zhe Song, and Zhanyu Ma

    Ruoyi Du, Dongliang Chang, Timothy M. Hospedales, Yi-Zhe Song, and Zhanyu Ma. 2024. DemoFusion: Democratising High-Resolution Image Generation With No $$$. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, W A, USA, June 16-22, 2024. IEEE, 6159–6168. doi:10.1109/CVPR52733. 2024.00589

  6. [7]

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas Müller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, Dustin Podell, Tim Dockhorn, Zion English, and Robin Rombach. 2024. Scaling Rectified Flow Transformers for High-Resolution Image Synthesis. In Forty-first International Conference on Machine Learning, ICML 2024...

  7. [8]

    Flux. [n. d.]. A Better Way to Build PCBs. https://www.flux.ai/

  8. [9]

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. 2020. Denoising Diffusion Probabilistic Models. arXiv:2006.11239 [cs.LG] https://arxiv.org/abs/2006.11239

Show all 28 references
  1. [10]

    Jonathan Ho and Tim Salimans. 2022. Classifier-Free Diffusion Guidance. CoRR abs/2207.12598 (2022). doi:10.48550/ARXIV.2207.12598 arXiv:2207.12598

  2. [12]

    JaidedAI. 2021. EasyOCR: A Ready-to-Use OCR with 80+ Languages. https: //github.com/JaidedAI/EasyOCR GitHub repository

  3. [13]

    Chao Li, Chen Jiang, Xiaolong Liu, Jun Zhao, and Guoxin Wang. 2025. JoyType: A Robust Design for Multilingual Visual Text Creation. arXiv:2409.17524 [cs.CV] https://arxiv.org/abs/2409.17524

  4. [14]

    Zeyu Liu, Weicong Liang, Zhanhao Liang, Chong Luo, Ji Li, Gao Huang, and Yuhui Yuan. 2024. Glyph-ByT5: A Customized Text Encoder for Accurate Visual Text Rendering. In Computer Vision - ECCV 2024 - 18th European Conference, Milan, Italy, September 29-October 4, 2024, Proceedin...

  5. [15]

    Jian Ma, Yonglin Deng, Chen Chen, Haonan Lu, and Zhenyu Yang. 2024. Glyph- Draw2: Automatic Generation of Complex Glyph Posters with Diffusion Models and Large Language Models. CoRR abs/2407.02252 (2024). doi:10.48550/ARXIV. 2407.02252 arXiv:2407.02252

  6. [16]

    Jian Ma, Mingjun Zhao, Chen Chen, Ruichen Wang, Di Niu, Haonan Lu, and Xiaodong Lin. 2023. GlyphDraw: Seamlessly Rendering Text with Intricate Spatial Structures in Text-to-Image Generation. arXiv:2303.17870 [cs.CV] https: //arxiv.org/abs/2303.17870

  7. [18]

    Chong Mou, Xintao Wang, Liangbin Xie, Yanze Wu, Jian Zhang, Zhongang Qi, and Ying Shan. 2024. T2I-Adapter: Learning Adapters to Dig Out More Controllable Ability for Text-to-Image Diffusion Models. In Thirty-Eighth AAAI Conference on Artificial Intelligence, AAAI 2024, Thirty-...

  8. [19]

    Dustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas Müller, Joe Penna, and Robin Rombach. 2024. SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis. In The Twelfth Interna- tional Conference on Learning Representations, ICLR ...

  9. [20]

    Yuxiang Tuo, Yifeng Geng, and Liefeng Bo. 2024. AnyText2: Visual Text Gen- eration and Editing With Customizable Attributes. arXiv:2411.15245 [cs.CV] https://arxiv.org/abs/2411.15245

  10. [21]

    Yuxiang Tuo, Wangmeng Xiang, Jun-Yan He, Yifeng Geng, and Xuansong Xie

  11. [22]

    Xiaoshi Wu, Yiming Hao, Keqiang Sun, Yixiong Chen, Feng Zhu, Rui Zhao, and Hongsheng Li. 2023. Human Preference Score v2: A Solid Benchmark for Evalu- ating Human Preferences of Text-to-Image Synthesis. arXiv:2306.09341 [cs.CV] https://arxiv.org/abs/2306.09341

  12. [23]

    Yukang Yang, Dongnan Gui, Yuhui Yuan, Weicong Liang, Haisong Ding, Han Hu, and Kai Chen. 2023. GlyphControl: Glyph Conditional Con- trol for Visual Text Generation. In Advances in Neural Information Process- ing Systems 36: Annual Conference on Neural Information Processing Sy...

  13. [24]

    In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024

    AnyText: Multilingual Visual Text Generation and Editing. In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024. OpenReview.net. https://openreview.net/forum?id=ezBH9WE9s2

  14. [25]

    Lingjun Zhang, Xinyuan Chen, Yaohui Wang, Yue Lu, and Yu Qiao. 2024. Brush Your Text: Synthesize Any Scene Text on Images via Diffusion Model. InThirty- Eighth AAAI Conference on Artificial Intelligence, AAAI 2024, Thirty-Sixth Con- ference on Innovative Applications of Artifi...

  15. [26]

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. 2023. Adding Conditional Control to Text-to-Image Diffusion Models. In IEEE/CVF International Conference on Computer Vision, ICCV 2023, Paris, France, October 1-6, 2023 . IEEE, 3813–3824. doi:10.1109/ICCV51070.2023.00355

  16. [27]

    Hu Ye, Jun Zhang, Sibo Liu, Xiao Han, and Wei Yang. 2023. IP-Adapter: Text Compatible Image Prompt Adapter for Text-to-Image Diffusion Models. CoRR abs/2308.06721 (2023). doi:10.48550/ARXIV.2308.06721 arXiv:2308.06721

  17. [28]

    ViT-B/32

    Yiming Zhao and Zhouhui Lian. 2024. UDiffText: A Unified Framework for High- Quality Text Synthesis in Arbitrary Images via Character-Aware Diffusion Models. In Computer Vision - ECCV 2024 - 18th European Conference, Milan, Italy, September 29-October 4, 2024, Proceedings, Par...

  18. [30]

    Metaxas, and Praveen Krishnan

    Qilong Zhangli, Jindong Jiang, Di Liu, Licheng Yu, Xiaoliang Dai, Ankit Ram- chandani, Guan Pang, Dimitris N. Metaxas, and Praveen Krishnan. 2024. Layout- Agnostic Scene Text Image Synthesis with Diffusion Models. In IEEE/CVF Confer- ence on Computer Vision and Pattern Recogni...

  19. [2023]

    TextDiffuser: Diffusion Models as Text Painters. In Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023, Alice Oh, Tristan Naumann, Amir Globerson, Kat...

  20. [2024]

    In Computer Vision - ECCV 2024 - 18th European Conference, Milan, Italy, Sep- tember 29-October 4, 2024, Proceedings, Part V (Lecture Notes in Computer Science, Vol

    TextDiffuser-2: Unleashing the Power of Language Models for Text Render- ing. In Computer Vision - ECCV 2024 - 18th European Conference, Milan, Italy, Sep- tember 29-October 4, 2024, Proceedings, Part V (Lecture Notes in Computer Science, Vol. 15063), Ales Leonardis, Elisa Ric...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.