Pith. sign in

REVIEW 4 major objections 5 minor 2 cited by

AMO Sampler: Enhancing Text Rendering with Overshooting

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper shows that a training-free overshooting sampler, alternating an over-simulated ODE step with matched noise injection, acts as a Langevin correction and improves text rendering accuracy on SD3, Flux, and AuraFlow without extra…

desk verdict A clever and correctly derived stochastic sampler for rectified flow whose headline Flux improvement doesn't match any reported number — a real but fixable problem. read the letter →

arxiv 2411.19415 v2 pith:B6J2C5OE submitted 2024-11-28 cs.CV cs.AIcs.LG

Xixi Hu , Keyang Xu , Bo Liu , Qiang Liu , Hongliang Fei This is my paper · ORCID
classification cs.CVcs.AIcs.LG
keywords textrenderingrectifiedflowtraining-freesamplingLangevindynamicsattentionmodulationtext-to-imagegenerationstochasticsamplerovershooting
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 tries to fix misspelled or garbled text in images produced by rectified-flow text-to-image models, without retraining or extra inference cost. It claims that alternating an overshot ODE step with a matched noise-injection step is mathematically equivalent to adding a Langevin dynamics term that pushes samples back toward the correct marginal distribution, countering errors accumulated by successive Euler steps. To avoid over-smoothing, it modulates the overshooting strength per image patch using cross-attention between text tokens and image patches, calling the combined sampler AMO. On SD3, Flux, and AuraFlow, the reported correct-text rate rises by 32.3% and 35.9% relative on the two stronger models, with FID and CLIP scores staying roughly level. A reader should care because this type of fix can be dropped into any pretrained flow model rather than requiring new training.

What carries the argument

The central object is the overshooting sampler update: from $\tilde Z_t$ at time $t$, compute $\hat Z_o = \tilde Z_t + (o-t)v(\tilde Z_t,t)$ with $o = t + (1+c)\epsilon$, then set $\tilde Z_s = a\hat Z_o + b\xi$ for $\xi\sim N(0,I)$, with $a = s/o$ and $b = \sqrt{(1-s)^2 - a^2(1-o)^2}$. In the $\epsilon\to0$ limit this becomes the SDE in the core discovery, i.e. the learned ODE plus a Langevin step controlled by $c$. The attention modulation replaces the scalar $c$ with $c\,m_{h,w}$, where $m_{h,w}$ is a normalized cross-attention map between the text tokens and each image patch, so that overshooting acts strongly on text-related regions and weakly elsewhere.

What would settle it

Compare AMO at $c=2$ with Euler on a rectified-flow model whose velocity has been deliberately biased in text regions, e.g. a model fine-tuned to produce misspelled text; if correct-text rates do not improve or worsen, the Langevin correction is chasing the wrong marginal. A more direct test on a synthetic flow with known $v$ would compare the empirical distribution of overshot samples $\hat Z_o$ to $oX_1+(1-o)X_0$ and reveal any mismatch in the coefficient formula $b=\sqrt{(1-s)^2-a^2(1-o)^2}$.

Watch

Extended reading notes

Core claim

The central claim is that the overshooting update, where one Euler step lands at time $o = s + c\epsilon$ and a coefficient-matched noise re-introduction returns to $s$, preserves the marginal law of the rectified flow in the small-step limit. In that limit the update becomes the SDE $dZ_t = ((1+c)v(Z_t,t) - c Z_t/t)\,dt + \sqrt{2c(1-t)/t}\,dW_t$, which equals the original ODE plus a Langevin correction term whose step size is $c(1-t)/t$. Because successive Euler steps make the particle distribution drift from $\rho_t$, the Langevin term acts as a denoising correction that improves text rendering; attention modulation then scales $c$ per patch using $m_{h,w}$, the normalized cross-attention weight between text tokens and image patch $(h,w)$, confining the correction to text areas. In the paper's own terms, 'AMO demonstrates a 32.3% and 35.9% improvement in text rendering accuracy on SD3 and Flux without compromising overall image quality or increasing inference cost.'

Load-bearing premise

The derivation assumes the learned velocity field matches the true conditional expectation, so that the overshot sample follows the correct flow marginal; in text regions, where the model is imperfect, this assumption is exactly what is least safe.

Editorial extensions

If this is right

  • On SD3, Flux, and AuraFlow, AMO raises the correct-text rate in human evaluation relative to Euler: 43.0% vs 32.5% on SD3, 82.5% vs 74.0% on Flux, and 3.0% vs 1.0% on AuraFlow.
  • The gains persist at 20, 50, and 100 sampling steps, with larger relative improvements at the smaller step counts.
  • FID and CLIP scores stay roughly level or improve slightly, so the text correction does not trade away overall image quality.
  • The sampler is training-free and model-agnostic for rectified flow: it only requires access to the velocity network and the cross-attention maps.
  • Ablations show both noise compensation and attention modulation are necessary; overshooting alone without the matched noise step collapses the marginal and yields a 0% correct-text rate.

Reading between the lines

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

  • Beyond the paper: the Langevin-correction mechanism should apply to any ODE-based generative model, not just rectified-flow text rendering, because the derivation only requires the marginal-preservation identity and the score formula $\nabla\log\rho_t=(tv(x,t)-x)/(1-t)$.
  • Beyond the paper: the attention mask $m_{h,w}$ could be reused or approximated with a single attention layer, and AMO could be evaluated on a hand-keypoint benchmark to quantify the appendix's anecdotal improvement in hands and body structure.
  • Beyond the paper: if the dominant text error is model bias rather than discretization drift, the correction will be limited; a synthetic test with a deliberately mis-specified velocity would separate the two error sources.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes a training-free stochastic sampling modification for pretrained rectified flow text-to-image models. At each denoising step, the Euler update is replaced by an 'overshooting' step that advances the ODE further than the target time and then adds noise whose coefficients are chosen to preserve the marginal law. An attention-modulated variant (AMO) scales the overshooting strength per image patch using cross-attention with text tokens, in order to reduce over-smoothing outside text regions. The paper derives an SDE limit for the overshooting sampler, shows its equivalence to a Langevin-corrected Euler step, and reports human and OCR evaluations on SD3, Flux, and AuraFlow claiming substantial text-rendering gains without extra training or inference cost.

Significance. If the central claims hold, this would be a useful and cheap inference-time improvement for a real bottleneck in text-to-image generation, and the attention-modulation mechanism is a sensible way to localize stochastic correction. The paper has notable strengths: the SDE-limit derivation in Appendix A.1 is algebraically sound; the Fokker-Planck argument in Appendix A.2 connecting the overshooting sampler to a Langevin term is standard and correctly executed; the attention-weighted elementwise noise compensation is a neat practical trick; and the code is publicly released. However, the headline quantitative claim for Flux is not reproducible from the reported table, the human evaluation lacks uncertainty estimates and is too small to support the main Flux gap, and the overshooting strength is tuned per model on the same evaluation benchmarks. These issues are load-bearing for the abstract-level claims and must be fixed before the empirical conclusions can be accepted.

major comments (4)
  1. [Abstract; Table 2] The abstract states that AMO demonstrates a 35.9% improvement in text rendering accuracy on Flux, but no metric in Table 2 yields that number. The correct rate rises from 74.0% to 82.5%, which is an 11.5% relative improvement; OCR-A rises 21.7%; OCR-F rises 7.9%. If the 35.9% figure comes from an unreported subset, such as the 20-step runs in Figure 5, the exact metric and step count must be stated and the underlying numbers reported. As written, the paper's central quantitative claim for Flux is not verifiable from its own results.
  2. [Section 5, Table 2; Appendix A.3] The human correct-rate evaluation uses 100 prompts with at least two raters per image, but no confidence intervals, inter-annotator agreement, or per-prompt breakdown are provided. For Flux, the 8.5-percentage-point gap between 74.0% and 82.5% has an approximate standard error of 5.8 percentage points under independent binomial sampling with n=100, so the difference is not statistically significant at the 5% level. The paper should report interval estimates and agreement statistics, and soften the claim of 'notably surpassing' Euler accordingly.
  3. [Section 5.2, Table 4, Figure 7] The overshooting strength c is a per-model hyperparameter (c=2 for Flux, c=1 for SD3 and AuraFlow in Table 4) and is selected on the same benchmark set used for evaluation: Figure 7 shows CR, OCR-F, and OCR-A versus c on Flux and the text says c=2 is chosen because performance plateaus there. This creates a selection-on-the-evaluation-set concern for the reported gains. The authors should either tune c on a separate validation split and report held-out numbers, or demonstrate that the qualitative conclusions are insensitive to c over a range fixed before evaluation.
  4. [Section 3.1, Eq. (3); Appendix A.2, Lemma A.2] The marginal-preservation guarantee and the score identity in Eq. (23) assume that the learned velocity v(Z_t,t) equals the true conditional expectation E[X_1-X_0 | X_t]. The paper's motivation is precisely that text rendering is difficult because the model's velocity is imperfect in text regions; if the dominant failure is model bias rather than Euler discretization error, the Langevin correction in Eq. (8) targets the wrong marginal and the coefficients a and b in Eq. (4) are not guaranteed to correct the actual error. This does not invalidate the empirical method, but the mechanism claim should be qualified and tested, for example by measuring velocity error or by comparing against a model with reliable text rendering.
minor comments (5)
  1. [Section 5.3 and Figure 8] The text refers to 'GraphControl', but the model and the cited reference are 'GlyphControl'.
  2. [Section 6] The conclusion contains the duplicated phrase 'Overshooting sampler sampler'.
  3. [Figure 5] The figure would be much more informative with numerical values and error bars; currently it only shows bars for three metrics at three step counts, and the text uses it to support the low-step-count advantage.
  4. [Appendix A.3 and Algorithm 1] The timestep schedule is not fully specified, in particular whether t0 is exactly 0 or a small epsilon; near t=0 the coefficients in Eq. (4) have a 1/o form, so the practical handling of the first step should be stated.
  5. [Algorithm 1, Eq. (9)] The attention mask m_i is listed as an input to the overshooting update, but the paper should clarify whether it reuses cross-attention activations from the velocity evaluation or requires an extra forward pass, since the claim of negligible overhead depends on this.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the overshooting sampler's marginal-preservation property is derived from the rectified-flow interpolation law, and the reported text-rendering gains are external empirical results, not fitted constants or self-citational restatements.

full rationale

The central derivation in Sections 3.1 and 3.2 is self-contained. The coefficients a and b are solved by matching the law of a(oX1+(1-o)X0)+bξ to sX1+(1-s)X0, which is a direct algebraic consequence of the rectified-flow interpolation Xt=tX1+(1-t)X0, not an assumption of the conclusion. The SDE limit and the equivalence to Langevin dynamics follow from the Fokker-Planck equation with σ_t^2=2c(1-t)/t, and Lemma A.2 derives ∇logρ_t(x)=(tv(x,t)-x)/(1-t) from E[X1-X0|Xt=x], the defining regression property of the learned velocity. The attention modulation mask is taken from the pretrained model's cross-attention, so it is an input to the sampler rather than a quantity that encodes the claimed improvement. The overshooting strength c is a tunable hyperparameter, and the paper's core claim does not reduce to fitting c to the reported correct rates; the default values c=2 (Flux) and c=1 (SD3/AuraFlow) are set before the comparison and the mechanism is stated independently of any particular c. The only self-citations are background references to rectified-flow literature by the same author group, and none is used as load-bearing evidence for the sampler's correctness or for the empirical gains. The abstract's 35.9% Flux improvement is not directly reproducible from Table 2, since 74.0% to 82.5% is an 11.5% relative increase; however, that is a reporting and statistical-verifiability concern, not circularity. The derivation does not reduce, by construction or by self-citation, to its own inputs.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The method introduces no new physical entities or model components; it is a sampling algorithm. The main added degrees of freedom are the per-model overshooting strength c and the assumption that attention scores correspond to text regions. The velocity-accuracy assumption is inherited from rectified flow training and is already known to fail for text rendering, which is the problem the paper tries to fix.

free parameters (1)
  • Overshooting strength c = 2.0 (Flux), 1.0 (SD3, AuraFlow)
    Selected per model via ablation on the same benchmarks used for evaluation (Section 5.2, Figure 7, Table 4).
assumptions (3)
  • domain assumption The learned velocity field approximates the true conditional expectation v(x_t,t) = E[X1 - X0 | X_t = x_t].
    Used in Lemma A.2 to identify the score function grad log rho_t and in Section 3.1 to assert the overshooting step preserves the marginal distribution. If v is biased, as it is precisely for text rendering, the correction term targets the wrong marginal.
  • standard math The latent variable X0 is standard Gaussian at t=0.
    Standard rectified flow assumption stated in Section 2.
  • domain assumption Cross-attention scores between text tokens and image patches localize the regions where text is being rendered.
    The attention modulation (Section 3.3, Eq. 9) relies on this to concentrate overshooting on text areas; if attention is diffuse or misaligned, the method's text gains and image-quality preservation break down.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AMO Sampler: Enhancing Text Rendering with Overshooting." pith.science (2026). https://pith.science/paper/B6J2C5OE

@misc{pith2026241119415,
  author       = {Pith},
  title        = {Pith review of: AMO Sampler: Enhancing Text Rendering with Overshooting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B6J2C5OE}},
  note         = {Machine review of arXiv:2411.19415}
}
read the original abstract

Achieving precise alignment between textual instructions and generated images in text-to-image generation is a significant challenge, particularly in rendering written text within images. Sate-of-the-art models like Stable Diffusion 3 (SD3), Flux, and AuraFlow still struggle with accurate text depiction, resulting in misspelled or inconsistent text. We introduce a training-free method with minimal computational overhead that significantly enhances text rendering quality. Specifically, we introduce an overshooting sampler for pretrained rectified flow (RF) models, by alternating between over-simulating the learned ordinary differential equation (ODE) and reintroducing noise. Compared to the Euler sampler, the overshooting sampler effectively introduces an extra Langevin dynamics term that can help correct the compounding error from successive Euler steps and therefore improve the text rendering. However, when the overshooting strength is high, we observe over-smoothing artifacts on the generated images. To address this issue, we propose an Attention Modulated Overshooting sampler (AMO), which adaptively controls the strength of overshooting for each image patch according to their attention score with the text content. AMO demonstrates a 32.3% and 35.9% improvement in text rendering accuracy on SD3 and Flux without compromising overall image quality or increasing inference cost. Code available at: https://github.com/hxixixh/amo-release.

Figures

Figures reproduced from arXiv: 2411.19415 by the authors.

Figure 1
Figure 1. Improved Text Rendering. (a)-(d) illustrate four common text rendering mistakes in text-to-image generations. Compared to the standard Euler sampler (purple), our Attention Modulated Overshooting sampler (AMO) (yellow) produces accurate and complete text without additional training, and remains as computationally efficient as the Euler sampler. Abstract Achieving precise alignment between textual instructions and ge… view at source ↗
Figure 2
Figure 2. Visualization of the Overshooting Sampler. Given [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Euler versus Overshooting on a toy dataset. The noise (π0) and data (π1) distributions are shown as blue and light-purple dots. Top: The samples from Euler deviate from π1. Overshooting sampler helps correct the marginal. As c increases, the correction effect is stronger, but it also introduces smoothing artifacts. Bottom: Starting with Z˜t (t = 0.5) from the Euler sampler, if we apply 5 times of (Overshooting - Eul… view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Comparison of text rendering quality between Euler and our stochastic sampling method across three different text-to-image models: (a) Flux, (b) Stable Diffusion 3 (SD3), and (c) AuraFlow. All results are generated using the same random seed for consistent comparison. …
Figure 5
Figure 5. Figure 5: The comparison of Euler sampler and AMO across dif [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6 [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Impact of overshooting strength on Text Rendering Performance. This figure illustrates how varying the overshooting strength parameter c in AMO affects the Flux model’s text render￾ing performance. Larger c tends to achieve higher text rendering quality. We observe tha…
Figure 8
Figure 8. Figure 8: Results of human evaluation comparing text rendering quality and overall image quality across five methods. Partici￾pants viewed five images, each generated by one of the methods, and were asked to vote for: (1) the models with the best text ren￾dering quality (multipl…
Figure 9
Figure 9. Figure 9: Examples of OCR model performance. Detected text boxes and prediction results are shown in red. The OCR model fails to detect [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]
Figure 10
Figure 10. Figure 10: Image Quality for Euler, Overshooting, and AMO. Please zoom in for finer details. The Overshooting method shown here employs a one-step overshooting strategy, ensuring the overall computational cost remains comparable across all three methods. The overshooting approac…
Figure 11
Figure 11. Figure 11: Samples generated by varying c. As c increases, the images gradually lose complexity and fine details due to over-smoothing. For moderate values of c, such as c = 2, the results achieve a balance between accurate text rendering and visual quality. A.5.2 Quantative Res…
Figure 12
Figure 12. Figure 12: Comparison of text rendering quality between Euler and AMO. Results are presented across three different text-to-image models: Flux, Stable Diffusion 3, and AuraFlow. All images are generated using the same random seed. In each pair of images, the left column shows th…
Figure 13
Figure 13. Figure 13: Comparison of samples generated by different methods, including TextDiffuser, GlyphControl, Euler, Overshooting, and AMO. During the human evaluation, participants were shown five images for comparison. A.7. Exploring Tasks Beyond Text Rendering Our initial exploratio…
Figure 14
Figure 14. Figure 14: Correcting hands and body structure using our method. [PITH_FULL_IMAGE:figures/full_fig_p018_14.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Momentum Guidance: Plug-and-Play Guidance for Flow Models

    cs.LG 2026-02 conditional novelty 6.0 of 10

    Momentum Guidance improves flow-model sample quality by extrapolating the current velocity away from an exponential moving average of past velocities, with no extra model evaluations.

  2. DreamActor-H1: High-Fidelity Human-Product Demonstration Video Generation via Motion-designed Diffusion Transformers

    cs.CV 2025-06 conditional novelty 6.0 of 10

    A diffusion transformer model generates human-product demonstration videos from paired human and product images while preserving both identities through masked cross-attention and motion template guidance.

Reference graph

Works this paper leans on

35 extracted references · 16 canonical work pages · cited by 2 Pith papers

  1. [1]

    Building nor- malizing flows with stochastic interpolants

    Michael S Albergo and Eric Vanden-Eijnden. Building nor- malizing flows with stochastic interpolants. arXiv preprint arXiv:2209.15571, 2022. 5

    arXiv 2022

  2. [2]

    Imagen 3

    Jason Baldridge, Jakob Bauer, Mukul Bhutani, Nicole Brich- tova, Andrew Bunner, Kelvin Chan, Yichang Chen, Sander Dieleman, Yuqing Du, Zach Eaton-Rosen, et al. Imagen 3. arXiv preprint arXiv:2408.07009, 2024. 5

    arXiv 2024

  3. [3]

    Improving image generation with better captions

    James Betker, Gabriel Goh, Li Jing, Tim Brooks, Jianfeng Wang, Linjie Li, Long Ouyang, Juntang Zhuang, Joyce Lee, Yufei Guo, et al. Improving image generation with better captions. Computer Science. https://cdn. openai. com/papers/dall-e-3. pdf, 2(3):8, 2023. 5

    work page 2023

  4. [4]

    Textdiffuser-2: Unleashing the power of language models for text rendering

    Jingye Chen, Yupan Huang, Tengchao Lv, Lei Cui, Qifeng Chen, and Furu Wei. Textdiffuser-2: Unleashing the power of language models for text rendering. arXiv preprint arXiv:2311.16465, 2023. 2, 5, 6

    arXiv 2023

  5. [5]

    Textdiffuser: Diffusion models as text painters

    Jingye Chen, Yupan Huang, Tengchao Lv, Lei Cui, Qifeng Chen, and Furu Wei. Textdiffuser: Diffusion models as text painters. Advances in Neural Information Processing Systems, 36, 2024. 2, 5, 8

    work page 2024

  6. [6]

    Diffusion models beat gans on image synthesis

    Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis. Advances in neural information processing systems, 34:8780–8794, 2021. 5

    work page 2021

  7. [7]

    Scaling rectified flow trans- formers for high-resolution image synthesis

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim En- tezari, Jonas Müller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, et al. Scaling rectified flow trans- formers for high-resolution image synthesis. In Forty-first International Conference on Machine Learning, 2024. 2, 5

    work page 2024

  8. [8]

    Aura Flow Pipeline Documentation , 2023

    Hugging Face. Aura Flow Pipeline Documentation , 2023. Huggingface. 5

    work page 2023

Show all 35 references
  1. [9]

    Flux that plays music

    Zhengcong Fei, Mingyuan Fan, Changqian Yu, and Jun- shi Huang. Flux that plays music. arXiv preprint arXiv:2409.00587, 2024. 5

    2024 arXiv

  2. [10]

    Iterative α-(de) blending: A minimalist deterministic diffusion model

    Eric Heitz, Laurent Belcour, and Thomas Chambon. Iterative α-(de) blending: A minimalist deterministic diffusion model. In ACM SIGGRAPH 2023 Conference Proceedings, pages 1–8, 2023. 5

    2023

  3. [11]

    Classifier-free diffusion guidance

    Jonathan Ho and Tim Salimans. Classifier-free diffusion guidance. arXiv preprint arXiv:2207.12598, 2022. 5

    2022 arXiv

  4. [12]

    Denoising diffu- sion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffu- sion probabilistic models. Advances in neural information processing systems, 33:6840–6851, 2020. 5

    2020

  5. [13]

    Video diffusion models

    Jonathan Ho, Tim Salimans, Alexey Gritsenko, William Chan, Mohammad Norouzi, and David J Fleet. Video diffusion models. Advances in Neural Information Processing Systems, 35:8633–8646, 2022. 5

    2022

  6. [14]

    Elucidating the design space of diffusion-based generative models

    Tero Karras, Miika Aittala, Timo Aila, and Samuli Laine. Elucidating the design space of diffusion-based generative models. Advances in neural information processing systems, 35:26565–26577, 2022. 4, 5

    2022

  7. [15]

    Black Forest Labs. Flux. https://github.com/ black-forest-labs/flux , 2023. GitHub repository. 5

    2023

  8. [16]

    V oicebox: Text-guided multilin- gual universal speech generation at scale

    Matthew Le, Apoorv Vyas, Bowen Shi, Brian Karrer, Leda Sari, Rashel Moritz, Mary Williamson, Vimal Manohar, Yossi Adi, Jay Mahadeokar, et al. V oicebox: Text-guided multilin- gual universal speech generation at scale. Advances in neural information processing systems, 36, 2024. 5

    2024

  9. [17]

    Mask textspotter v3: Segmentation proposal network for robust scene text spotting

    Minghui Liao, Guan Pang, Jing Huang, Tal Hassner, and Xi- ang Bai. Mask textspotter v3: Segmentation proposal network for robust scene text spotting. In Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23– 28, 2020, Proceedings, Part XI 16, pages 706–72...

    2020

  10. [18]

    Flow matching for generative modeling

    Yaron Lipman, Ricky TQ Chen, Heli Ben-Hamu, Maximilian Nickel, and Matt Le. Flow matching for generative modeling. arXiv preprint arXiv:2210.02747, 2022. 2, 5

    2022 arXiv

  11. [19]

    Rectified flow: A marginal preserving approach to optimal transport

    Qiang Liu. Rectified flow: A marginal preserving approach to optimal transport. arXiv preprint arXiv:2209.14577, 2022. 2

    2022 arXiv

  12. [20]

    Character-aware models improve visual text rendering

    Rosanne Liu, Dan Garrette, Chitwan Saharia, William Chan, Adam Roberts, Sharan Narang, Irina Blok, RJ Mical, Mo- hammad Norouzi, and Noah Constant. Character-aware models improve visual text rendering. arXiv preprint arXiv:2212.10562, 2022. 5

    2022 arXiv

  13. [21]

    Flow straight and fast: Learning to generate and transfer data with rectified flow

    Xingchao Liu, Chengyue Gong, and Qiang Liu. Flow straight and fast: Learning to generate and transfer data with rectified flow. In The Eleventh International Conference on Learning Representations, 2022. 5

    2022

  14. [22]

    Glyph-byt5-v2: A strong aesthetic baseline for accurate multilingual visual text rendering

    Zeyu Liu, Weicong Liang, Yiming Zhao, Bohan Chen, Ji Li, and Yuhui Yuan. Glyph-byt5-v2: A strong aesthetic baseline for accurate multilingual visual text rendering. arXiv preprint arXiv:2406.10208, 2024. 5, 8

    2024 arXiv

  15. [23]

    Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps

    Cheng Lu, Yuhao Zhou, Fan Bao, Jianfei Chen, Chongxuan Li, and Jun Zhu. Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps. Advances in Neural Information Processing Systems, 35:5775–5787, 2022. 5

    2022

  16. [24]

    Glyphdraw: Seamlessly rendering text with intricate spatial structures in text-to-image generation

    Jian Ma, Mingjun Zhao, Chen Chen, Ruichen Wang, Di Niu, Haonan Lu, and Xiaodong Lin. Glyphdraw: Seamlessly rendering text with intricate spatial structures in text-to-image generation. arXiv preprint arXiv:2303.17870, 2023. 5

    2023 arXiv

  17. [25]

    On distillation of guided diffusion models

    Chenlin Meng, Robin Rombach, Ruiqi Gao, Diederik Kingma, Stefano Ermon, Jonathan Ho, and Tim Salimans. On distillation of guided diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14297–14306, 2023. 5

    2023

  18. [26]

    Sdxl: Improving latent diffusion models for high-resolution image synthesis

    Dustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas Müller, Joe Penna, and Robin Rombach. Sdxl: Improving latent diffusion models for high-resolution image synthesis. arXiv preprint arXiv:2307.01952, 2023. 5

    2023 arXiv

  19. [27]

    Movie gen: A cast of media foundation models

    Adam Polyak, Amit Zohar, Andrew Brown, Andros Tjandra, Animesh Sinha, Ann Lee, Apoorv Vyas, Bowen Shi, Chih- Yao Ma, Ching-Yao Chuang, et al. Movie gen: A cast of media foundation models. arXiv preprint arXiv:2410.13720,

    arXiv

  20. [28]

    Learning transferable visual models from natural language supervi- sion

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervi- sion. In International conference on machine learning, ...

    2021

  21. [29]

    Hierarchical text-conditional image genera- tion with clip latents

    Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. Hierarchical text-conditional image genera- tion with clip latents. arXiv preprint arXiv:2204.06125, 1(2): 3, 2022. 5

    2022 arXiv

  22. [30]

    Pho- torealistic text-to-image diffusion models with deep language understanding

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily L Denton, Kamyar Ghasemipour, Raphael Gontijo Lopes, Burcu Karagol Ayan, Tim Salimans, et al. Pho- torealistic text-to-image diffusion models with deep language understanding. Advances in neural informati...

    2022

  23. [31]

    Denoising diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502,

    2010 arXiv

  24. [32]

    Generative modeling by estimating gradients of the data distribution

    Yang Song and Stefano Ermon. Generative modeling by estimating gradients of the data distribution. Advances in neural information processing systems, 32, 2019. 5

    2019

  25. [33]

    Score-based generative modeling through stochastic differential equations

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Ab- hishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equations. arXiv preprint arXiv:2011.13456, 2020. 2, 3, 5

    2011 arXiv

  26. [34]

    Anytext: Multilingual visual text gener- ation and editing

    Yuxiang Tuo, Wangmeng Xiang, Jun-Yan He, Yifeng Geng, and Xuansong Xie. Anytext: Multilingual visual text gener- ation and editing. arXiv preprint arXiv:2311.03054, 2023. 5

    2023 arXiv

  27. [35]

    Game Over

    Yiming Zhao and Zhouhui Lian. Udifftext: A unified framework for high-quality text synthesis in arbitrary im- ages via character-aware diffusion models. arXiv preprint arXiv:2312.04884, 2023. 5 A. Appendix A.1. The SDE Limit of the Overshooting Sampler In this section, we deri...

    2023 arXiv

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.