Pith. sign in

REVIEW 4 major objections 4 minor 38 references

Reusing Computation in Text-to-Image Diffusion for Efficient Generation of Image Sets

T0 review · 4 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Related text-to-image prompts can share early denoising steps, cutting compute by 50-74% while keeping quality.

desk verdict A clever, training-free compute-sharing trick that deserves peer review, but the 50–74% savings claim doesn't hold up until the authors say which diffusion stage they're accelerating and measure end-to-end. read the letter →

arxiv 2508.21032 v1 pith:II4F6ISW submitted 2025-08-28 cs.CV

Dale Decatur , Thibault Groueix , Wang Yifan , Rana Hanocka , Vladimir Kim , Matheus Gadelha This is my paper
classification cs.CV
keywords text-to-imagediffusioncomputereusepromptclusteringhierarchicalembeddingtreeimage-setgenerationUnCLIPpriorVQAscoretraining-freeacceleration
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

This paper tries to establish that generating sets of related text-to-image prompts can be made about two to four times cheaper if the diffusion model shares early denoising steps across prompts whose embeddings are similar. The authors argue that models trained with an UnCLIP-style image-embedding prior produce coarse layout before fine details, so early steps carry mostly shared, low-frequency content that a mean embedding can represent without losing per-prompt identity. Their training-free, clustering-based method reports compute savings of 50-74% on four prompt collections, with VQA-based quality comparable to or better than standard 40-step diffusion. If true, it gives an orthogonal efficiency axis to existing per-image fast-sampling and distillation, one that improves as the set of prompts grows.

What carries the argument

The load-bearing object is the hierarchical embedding tree: leaves are per-prompt text embeddings, internal nodes store the mean embedding of their children and a heterogeneity score cscore equal to the cosine distance between the two merged child clusters. At step k, phi(k)=tau(1-k/K) sets an allowed-heterogeneity threshold that falls linearly; each prompt uses the deepest node whose parent still clears the threshold, and the denoiser is evaluated once per selected node. Algorithm 1 copies the parent latent at the branch (line 15) and switches conditioning to the child embedding—the step that creates the savings, and also the step whose validity depends on the model's training prior.

What would settle it

Run Algorithm 1 with K=40 and tau=1 on two textually close but attribute-opposite prompts ('a cat wearing a hat' vs 'a cat without a hat') with Kandinsky, both starting from the same parent latent at the split step. If the shared-parent initialization consistently locks both images into the same hat/no-hat state or drops per-prompt VQA correctness below the standard two-run baseline across many pairs, the claim that a copied parent latent can be specialized by later steps is wrong.

Watch

Extended reading notes

Core claim

Central claim: in diffusion models trained with an UnCLIP-style image-embedding prior (Kandinsky, Karlo), fine detail emerges late, so early denoising steps for similar prompts are nearly interchangeable. The paper builds an agglomerative clustering tree over prompt embeddings and runs early steps on cluster-mean embeddings, branching to per-prompt embeddings once a heterogeneity threshold is crossed. On four datasets it reports 50-74% step savings at equal or better VQA Score; at roughly 31 shared steps per image it beats standard 40-step diffusion. The qualifier: models without the prior (Stable Diffusion, Stable UnCLIP) lock structure too early and show much less benefit.

Load-bearing premise

The entire saving hinges on the assumption that the partially denoised latent produced under a cluster's averaged embedding remains a valid starting point for each child prompt's remaining denoising steps; if copying the parent latent pushes a child off its own generation manifold, the shared steps corrupt rather than accelerate the image.

Editorial extensions

If this is right

  • A single cluster-mean trajectory replaces many near-identical early trajectories, moving the cost curve for N related prompts from roughly N times K denoiser calls toward K times log N calls on a balanced tree.
  • At equal quality, total denoising steps drop to 26-50% of standard inference for Kandinsky and Karlo; at equal step counts, VQA quality is higher than standard diffusion.
  • Savings increase with set size and semantic homogeneity: 74% on 100 style variations, 76% on 500 subject variations, and 65.3% on 16 virtual try-ons.
  • The method is a wrapper around an existing denoiser: no training or fine-tuning, so it composes with any scheduler and can be combined with step-reduction tools.
  • The benefit is not universal: at comparable quality, SD 1.5 saves up to about 28% and FLUX up to about 24%, because structure emerges earlier in those models.

Reading between the lines

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

  • We infer the quality lift is not purely saved steps: sharing a cluster mean may act as implicit ensembling that stabilizes the early trajectory, which would explain why roughly 31 shared steps beat 40 independent steps on VQA; this is testable by comparing shared-mean and individual-prompt early steps at equal total FLOPs.
  • The cscore/phi scheduling suggests a portable diagnostic: measure how late high-frequency detail appears in any new model's intermediate latents (e.g., spectral energy) to predict sharing headroom before running a full set experiment.
  • For small, diverse sets the tree branches almost immediately, so the method should underperform there; an extension would decide cluster-by-cluster whether sharing is worthwhile instead of using one global tau.
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 / 4 minor

Summary. The paper proposes a training-free method to reduce computation when generating a set of images from multiple text prompts in text-to-image diffusion models. It constructs a hierarchical agglomerative tree over text embeddings, then shares early denoising steps by using averaged cluster embeddings as the conditioning signal, progressively specializing to individual prompt embeddings at later steps. Experiments on Kandinsky 2.2 and Karlo report 50–74% reductions in denoising steps relative to a 40-step baseline, with VQA-based win rates of 50.75–52.80% (i.e., approximate parity). The method is also evaluated on stable diffusion and FLUX, with smaller savings, and on several prompt-set applications (style variations, subject variations, virtual try-on).

Significance. If the central claims are established, the work offers a useful orthogonal efficiency direction for batch text-to-image generation, complementing per-inference acceleration methods. The idea is simple, training-free, and supported by pseudocode, ablations, and qualitative comparisons. The strongest part is the hierarchical sharing mechanism, which is clearly described and mechanically reduces the number of denoiser evaluations. However, the headline claims about 'computational cost' and 'improving image quality' require substantial qualification: the compute savings appear to be measured for a single diffusion stage (not the full pipeline for unCLIP-style models), and the quality evidence is near-parity without statistical guarantees. The paper does ship reproducible-style artifacts (pseudocode, project page) and includes several diverse datasets, which are strengths.

major comments (4)
  1. [§4, §5, Eq. (8), Table 1] It is unspecified which diffusion stage is shared for the unCLIP-style models Kandinsky and Karlo. These models comprise a text-to-image prior (text embedding → image embedding) and a latent diffusion decoder (image embedding → image). Eq. (8) writes the conditioning as \bar{e}_{f(y,k)}, but for the decoder the conditioning space is image embeddings, not text embeddings; for the prior it is text embeddings but the prior is a different diffusion process. The paper never states whether the reported savings apply to the prior, the decoder, or both, nor does it report the relative cost of these stages. Without this, the claim of saving '50% to 74% of the computational cost' (Table 1 caption) is not established for end-to-end text-to-image generation. Please specify the stage, provide an end-to-end FLOP or wall-clock comparison, and report the cost split.
  2. [§5.1, Table 1] The quality evidence is overclaimed. The win rates in Table 1 are 50.75%, 51.55%, 52.00%, and 52.80%—effectively parity—and no error bars, significance tests, or multiple-seed analysis are provided. The abstract's statement that the method 'significantly reduces compute cost while improving image quality' is not supported; the data support 'comparable quality' at best. Additionally, the compute-savings numbers are obtained by tuning τ per dataset to match the baseline VQA quality (as stated for Table 2), but the tuning procedure is not described. Please report confidence intervals or a proper statistical analysis, and temper the quality claim to 'comparable' unless stronger evidence is added.
  3. [Algorithm 1, Eq. (8), Fig. 4] The core mechanism assumes that a latent trajectory denoised with a cluster-mean embedding can be copied to each child prompt and then conditioned on the child's own embedding (Algorithm 1 lines 15–18). This is an empirical assumption: Fig. 4 shows it fails for Stable Diffusion and Stable UnCLIP, and the paper provides no quantitative analysis of when it succeeds. For the Kandinsky/Karlo cases the savings depend on this assumption holding. A short analysis—e.g., measuring latent divergence after switching conditioning, or a condition on the model's coarse-to-fine behavior—would make the method's scope and failure modes clearer. As written, the paper only offers a qualitative explanation.
  4. [Abstract, §1] The phrase 'for models trained conditioned on image embeddings, our approach significantly reduces compute cost while improving image quality' is misleading on two counts: (i) the compute savings are step-savings in a single diffusion process, not full-pipeline savings, and (ii) the quality improvement is not substantiated by the reported metrics. The claims should be aligned with the evidence: 'reduces denoising steps in the prior/decoder while maintaining comparable quality' would be accurate.
minor comments (4)
  1. [Fig. 5] The axes are ambiguous: please clarify what 'Compute Budget' includes (total denoiser evaluations? per-image average?) and define the unit (steps per image, total steps, FLOPs).
  2. [§5.1, Table 1] Please define exactly how 'Compute Saved' is computed, and state whether it refers to average denoiser evaluations per image relative to a 40-step standard run.
  3. [§1] The phrase 'as few as 26% of the total diffusion steps' is consistent with the 74% savings in Table 1, but it would be clearer to state that this is for the shared diffusion stage, not the full generation pipeline.
  4. [Supplementary, Algorithm 1] The timestep indexing in line 18 (using k−1) may be off by one relative to the main text's Eq. (8) (which uses t). Please check consistency.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: core claims are evaluated externally, and compute savings are a direct consequence of the algorithm rather than a fitted prediction.

full rationale

The paper's central claim is that sharing early denoising steps across semantically similar prompts via a hierarchical embedding tree saves 50–74% of diffusion steps while preserving or improving visual quality. Walking the derivation chain, no step reduces to its own inputs by construction. The method (Sec. 4, Eq. 8, Algorithm 1) defines the sharing procedure, and compute savings follow arithmetically from performing fewer unique denoiser evaluations; the paper does not present this step-count reduction as an empirical prediction, so it is not circular. Quality is tested against the external VQA Score benchmark on GenAI Bench, Prompt Template, Style Variations, and Animals datasets, and the fixed-compute-budget comparison (Fig. 5) shows higher VQA scores than standard diffusion at equal budgets—an independent, non-forced result. The matched-quality numbers in Table 1 are obtained by tuning tau per dataset; this is a legitimate operating-point analysis rather than a fitted input called prediction, since the fixed-compute experiment does not rely on that tuning to show quality preservation. There are no load-bearing self-citations: the cited UnCLIP/Karlo/Kandinsky works are external, and no uniqueness theorem from the authors is invoked. The skeptic's concern that savings are reported for a diffusion stage rather than the full unCLIP pipeline is a scope/correctness limitation, not circularity. The paper's own Sec. 5.5 limitations do not assert any circular step. Verdict: no significant circularity; score 0.

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

The method depends on an untested interpolation assumption for averaged text embeddings, on coarse-to-fine behavior of specific model families, and on a tuned threshold tau. It introduces no free physical constants and no new entities.

free parameters (1)
  • tau = 1.0 default, sweep 0, 0.5, 1, 1.5
    Controls how many early denoising steps are shared. Chosen empirically to balance compute savings against VQA quality, and tuned to match baseline quality in the main savings experiments.
assumptions (3)
  • domain assumption Diffusion models generate images coarse-to-fine, with early steps defining low-frequency structure.
    Invoked in the Introduction and Fig. 2 to justify why early steps can be shared.
  • ad hoc to paper The mean of a cluster's text embeddings is a valid conditioning signal whose denoised latent remains compatible with every child prompt.
    Required by Eq. (8) and Algorithm 1. Empirically supported only for some models (Fig. 10), and fails for Stable Diffusion and Stable UnCLIP in Fig. 4.
  • domain assumption Models trained with a text-to-image prior exhibit sufficiently gradual detail emergence to leave room for shared early steps.
    The paper's own Fig. 2 and ablations show this is an empirical property of specific model families, not a general theorem.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Reusing Computation in Text-to-Image Diffusion for Efficient Generation of Image Sets." pith.science (2026). https://pith.science/paper/II4F6ISW

@misc{pith2026250821032,
  author       = {Pith},
  title        = {Pith review of: Reusing Computation in Text-to-Image Diffusion for Efficient Generation of Image Sets},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/II4F6ISW}},
  note         = {Machine review of arXiv:2508.21032}
}
read the original abstract

Text-to-image diffusion models enable high-quality image generation but are computationally expensive. While prior work optimizes per-inference efficiency, we explore an orthogonal approach: reducing redundancy across correlated prompts. Our method leverages the coarse-to-fine nature of diffusion models, where early denoising steps capture shared structures among similar prompts. We propose a training-free approach that clusters prompts based on semantic similarity and shares computation in early diffusion steps. Experiments show that for models trained conditioned on image embeddings, our approach significantly reduces compute cost while improving image quality. By leveraging UnClip's text-to-image prior, we enhance diffusion step allocation for greater efficiency. Our method seamlessly integrates with existing pipelines, scales with prompt sets, and reduces the environmental and financial burden of large-scale text-to-image generation. Project page: https://ddecatur.github.io/hierarchical-diffusion/

Figures

Figures reproduced from arXiv: 2508.21032 by the authors.

Figure 1
Figure 1. When generating a collection of images from text prompts, our approach shares denoising steps and can produce images of [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Coarse-to-Fine Generation. We show intermediate la￾tents from the diffusion process for Stable Diffusion 1.5 [28] (top row), Stable UnCLIP [26] (second row), Karlo [14] (third row), and Kandinsky [27] (bottom row). The models trained without a text-to-image prior (Stable Diffusion and Stable UnCLIP) learn structural details and high frequency features earlier on in the dif￾fusion process. In contrast, Karlo and Kand… view at source ↗
Figure 3
Figure 3. Tree Traversal. Left: our approach relies on a tree structure obtained by running agglomerative clustering on a set of prompt embeddings. Each node in the tree contains the average of the embeddings of its children, and a heterogeneity score c score based on the distance between its two children. To connect the tree hierarchy to the denoising steps, we design a function ϕ taking as input the denoising step k, and co… view at source ↗
Figures from the paper (7 more)
Figure 5
Figure 5. Figure 5: Generation Quality at Fixed Compute Budgets. We report VQA Score [17] on the Prompt Template Dataset [5] for both our method (orange) and the standard approach (blue) over various diffusion step budgets. Note that at a 40-step budget, our approach is identical to the s…
Figure 4
Figure 4. Figure 4: Importance of text-to-image prior. We share diffusion steps using a right-splitting binary tree as our hierarchy and the text prompts shown above for text-to-image diffusion models with various architectures. Kandinsky and Karlo are trained using an UnCLIP-style text-t…
Figure 6
Figure 6. Figure 6: Qualitative Comparison. We compare our approach (middle) with standard diffusion for a fixed compute budget of 18 steps (left) and a full compute budget of 40 steps (right). For the fixed compute budget, our method produces higher quality images than the standard appro…
Figure 7
Figure 7. Figure 7: Image Style Variations. We show an application of our method for efficiently generating style variations on given input prompt. We show a subset of 100 generated images, saving 74% of the equivalent compute for the standard approach. plates, including stylistic and str…
Figure 8
Figure 8. Figure 8: Virtual try-on. Our method generates a set of only 16 images, all containing the same subject, with various accessories, saving 65.3% of the equivalent compute for the standard approach. ‘‘An image of a <animal> wearing a birthday hat with a cake and birthday candles" …
Figure 9
Figure 9. Figure 9: Subject Variation. We show an application of our method for efficiently generating subject variations on given in￾put prompt. We show a subset of 500 generated images, saving 76% of the equivalent compute for the standard approach. computational efficiency even with re…
Figure 10
Figure 10. Figure 10: De-noised mean embedding (left) and its children using the Kandinsky model. [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

38 extracted references · 28 canonical work pages

  1. [1]

    Adobe firefly, 2023

    Adobe. Adobe firefly, 2023. Accessed: 2024. 1

    work page 2023

  2. [2]

    Prompting for discovery: Flexible sense-making for ai art- making with dreamsheets

    Shm Garanganao Almeda, JD Zamfirescu-Pereira, Kyu Won Kim, Pradeep Mani Rathnam, and Bjoern Hartmann. Prompting for discovery: Flexible sense-making for ai art- making with dreamsheets. In CHI Conference on Human Factors in Computing Systems, 2024. 1, 3

    work page 2024

  3. [3]

    Promptify: Text-to-image generation through interactive prompt exploration with large language models

    Stephen Brade, Bryan Wang, Mauricio Sousa, Sageev Oore, and Tovi Grossman. Promptify: Text-to-image generation through interactive prompt exploration with large language models. In Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology, 2023. 1, 3

    work page 2023

  4. [4]

    Large scale gan training for high fidelity natural image synthesis

    Andrew Brock, Jeff Donahue, and Karen Simonyan. Large scale gan training for high fidelity natural image synthesis. arXiv preprint arXiv:1809.11096, 2018. 2

    arXiv 2018

  5. [5]

    The prompt artists

    Minsuk Chang, Stefania Druga, Alexander J Fiannaca, Pedro Vergani, Chinmay Kulkarni, Carrie J Cai, and Michael Terry. The prompt artists. In Proceedings of the 15th Conference on Creativity and Cognition, pages 75–87, 2023. 3, 5, 6

    work page 2023

  6. [6]

    Prompt Expansion for Adaptive Text-to-Image Generation

    Siddhartha Datta, Alexander Ku, Deepak Ramachandran, and Peter Anderson. Prompt expansion for adaptive text-to- image generation. arXiv preprint arXiv:2312.16720, 2023. 1, 3

    work page Pith review arXiv 2023

  7. [7]

    Scaling recti- fied flow transformers for high-resolution image synthesis

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas M ¨uller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, et al. Scaling recti- fied flow transformers for high-resolution image synthesis. In Forty-first international conference on machine learning,

  8. [8]

    Generative adversarial nets

    Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. Advances in neural information processing systems, 27, 2014. 2

    2014

Show all 38 references
  1. [9]

    Denoising dif- fusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models. Advances in neural information processing systems, 33:6840–6851, 2020. 2, 3

    2020

  2. [10]

    Joe H. Ward Jr. Hierarchical grouping to optimize an objec- tive function. Journal of the American Statistical Associa- tion, 58(301):236–244, 1963. 4

    1963

  3. [11]

    Scal- ing up gans for text-to-image synthesis

    Minguk Kang, Jun-Yan Zhu, Richard Zhang, Jaesik Park, Eli Shechtman, Sylvain Paris, and Taesung Park. Scal- ing up gans for text-to-image synthesis. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10124–10134, 2023. 2

    2023

  4. [12]

    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. 3

    2022

  5. [13]

    Consistency trajectory mod- els: Learning probability flow ode trajectory of diffusion

    Dongjun Kim, Chieh-Hsin Lai, Wei-Hsiang Liao, Naoki Mu- rata, Yuhta Takida, Toshimitsu Uesaka, Yutong He, Yuki Mitsufuji, and Stefano Ermon. Consistency trajectory mod- els: Learning probability flow ode trajectory of diffusion. ICLR, 2024. 3

    2024

  6. [14]

    Karlo- v1.0.alpha on coyo-100m and cc15m.https://github

    Donghoon Lee, Jiseob Kim, Jisu Choi, Jongmin Kim, Min- woo Byeon, Woonhyuk Baek, and Saehoon Kim. Karlo- v1.0.alpha on coyo-100m and cc15m.https://github. com/kakaobrain/karlo, 2022. 2, 3

    2022

  7. [15]

    Genai-bench: Evaluating and improv- ing compositional text-to-visual generation

    Baiqi Li, Zhiqiu Lin, Deepak Pathak, Jiayao Li, Yixin Fei, Kewen Wu, Tiffany Ling, Xide Xia, Pengchuan Zhang, Gra- ham Neubig, et al. Genai-bench: Evaluating and improv- ing compositional text-to-visual generation. arXiv preprint arXiv:2406.13743, 2024. 5, 6

    2024 arXiv

  8. [16]

    Sdxl- lightning: Progressive adversarial diffusion distillation

    Shanchuan Lin, Anran Wang, and Xiao Yang. Sdxl- lightning: Progressive adversarial diffusion distillation. arXiv preprint arXiv:2402.13929, 2024. 1, 3

    2024 arXiv

  9. [17]

    Evaluating text-to-visual generation with image-to-text gen- eration

    Zhiqiu Lin, Deepak Pathak, Baiqi Li, Jiayao Li, Xide Xia, Graham Neubig, Pengchuan Zhang, and Deva Ramanan. Evaluating text-to-visual generation with image-to-text gen- eration. In European Conference on Computer Vision, pages 366–384. Springer, 2024. 5

    2024

  10. [18]

    Pseudo numerical methods for diffusion models on manifolds.arXiv preprint arXiv:2202.09778, 2022

    Luping Liu, Yi Ren, Zhijie Lin, and Zhou Zhao. Pseudo numerical methods for diffusion models on manifolds.arXiv preprint arXiv:2202.09778, 2022. 3

    2022 arXiv

  11. [19]

    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

    2022

  12. [20]

    Dpm-solver++: Fast solver for guided sampling of diffusion probabilistic models

    Cheng Lu, Yuhao Zhou, Fan Bao, Jianfei Chen, Chongx- uan Li, and Jun Zhu. Dpm-solver++: Fast solver for guided sampling of diffusion probabilistic models. arXiv preprint arXiv:2211.01095, 2022. 3

    2022 arXiv

  13. [21]

    Power hungry processing: Watts driving the cost of ai deployment? In Proceedings of the 2024 ACM conference on fairness, ac- countability, and transparency, 2024

    Sasha Luccioni, Yacine Jernite, and Emma Strubell. Power hungry processing: Watts driving the cost of ai deployment? In Proceedings of the 2024 ACM conference on fairness, ac- countability, and transparency, 2024. 1

    2024

  14. [22]

    Latent consistency models: Synthesizing high- resolution images with few-step inference

    Simian Luo, Yiqin Tan, Longbo Huang, Jian Li, and Hang Zhao. Latent consistency models: Synthesizing high- resolution images with few-step inference. ICLR, 2024. 2, 3

    2024

  15. [23]

    Is it ai or is it me? understanding users’ prompt journey with text-to-image generative ai tools

    Atefeh Mahdavi Goloujeh, Anne Sullivan, and Brian Magerko. Is it ai or is it me? understanding users’ prompt journey with text-to-image generative ai tools. In CHI Con- ference on Human Factors in Computing Systems, 2024. 3

    2024

  16. [24]

    Midjourney, 2022

    Midjourney, Inc. Midjourney, 2022. Accessed: 2024. 1

    2022

  17. [25]

    Dall-e, 2021

    OpenAI. Dall-e, 2021. Accessed: 2024. 1, 2

    2021

  18. [26]

    Hierarchical text-conditional image gener- ation with clip latents

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

    2022 arXiv

  19. [27]

    Kandinsky: an improved text-to-image syn- thesis with image prior and latent diffusion.EMNLP demoes,

    Anton Razzhigaev, Arseniy Shakhmatov, Anastasia Malt- seva, Vladimir Arkhipkin, Igor Pavlov, Ilya Ryabov, An- gelina Kuts, Alexander Panchenko, Andrey Kuznetsov, and Denis Dimitrov. Kandinsky: an improved text-to-image syn- thesis with image prior and latent diffusion.EMNLP demoes,

  20. [28]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022. 1, 2, 3, 8

    2022

  21. [29]

    Runway: Ai magic tools, 2022

    Runway AI. Runway: Ai magic tools, 2022. Accessed:

    2022

  22. [30]

    Photorealistic 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. Photorealistic text-to-image diffusion models with deep language understanding. Advances in neural information...

    2022

  23. [31]

    Progressive distillation for fast sampling of diffusion models

    Tim Salimans and Jonathan Ho. Progressive distillation for fast sampling of diffusion models. ICLR, 2022. 3

    2022

  24. [32]

    Deep unsupervised learning using nonequilibrium thermodynamics

    Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsupervised learning using nonequilibrium thermodynamics. In International confer- ence on machine learning, pages 2256–2265. pmlr, 2015. 2, 3

    2015

  25. [33]

    Denois- ing diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denois- ing diffusion implicit models. ICLR, 2021. 3

    2021

  26. [34]

    Visual autoregressive modeling: Scalable image generation via next-scale prediction

    Keyu Tian, Yi Jiang, Zehuan Yuan, Bingyue Peng, and Li- wei Wang. Visual autoregressive modeling: Scalable image generation via next-scale prediction. Advances in neural in- formation processing systems, 37:84839–84865, 2025. 8

    2025

  27. [35]

    Kandinsky 3: Text-to-image synthesis for multifunctional generative framework

    Arkhipkin Vladimir, Viacheslav Vasilev, Andrei Filatov, Igor Pavlov, Julia Agafonova, Nikolai Gerasimenko, Anna Averchenkova, Evelina Mironova, Bukashkin Anton, Kon- stantin Kulikov, Andrey Kuznetsov, and Denis Dimitrov. Kandinsky 3: Text-to-image synthesis for multifunctional...

    2024

  28. [36]

    One-step diffusion with distribution matching distillation

    Tianwei Yin, Micha ¨el Gharbi, Richard Zhang, Eli Shecht- man, Fredo Durand, William T Freeman, and Taesung Park. One-step diffusion with distribution matching distillation. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 6613–6623...

    2024

  29. [37]

    Improved distribution matching distillation for fast image synthesis

    Tianwei Yin, Micha ¨el Gharbi, Taesung Park, Richard Zhang, Eli Shechtman, Fredo Durand, and Bill Freeman. Improved distribution matching distillation for fast image synthesis. Advances in Neural Information Processing Systems , 37: 47455–47487, 2025. 3

    2025

  30. [38]

    mean image

    Wenliang Zhao, Lujia Bai, Yongming Rao, Jie Zhou, and Jiwen Lu. Unipc: A unified predictor-corrector framework for fast sampling of diffusion models. Advances in Neural Information Processing Systems, 36:49842–49869, 2023. 3 10 Reusing Computation in Text-to-Image Diffusion fo...

    2023

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.