MetricGold: Leveraging Text-To-Image Latent Diffusion Models for Metric Depth Estimation

Ansh Shah; K Madhava Krishna

REVIEW 3 major objections 5 minor 19 references

MetricGold: Leveraging Text-To-Image Latent Diffusion Models for Metric Depth Estimation

T0 review · 3 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 MetricGold claims that a pretrained text-to-image latent diffusion model can be fine-tuned on photorealistic synthetic RGB-D data—in about two days on a single consumer GPU—to produce metric depth maps that zero-shot transfer to real…

desk verdict A plausible latent-diffusion fine-tuning recipe for metric depth with synthetic-only data, but the preprint contains no experiments and the metric-scaling inverse is undefined. read the letter →

arxiv 2411.10886 v2 pith:CZTWRNXH submitted 2024-11-16 cs.CV cs.AIcs.GRcs.RO

Ansh Shah , K Madhava Krishna This is my paper · ORCID

classification cs.CVcs.AIcs.GRcs.RO

keywords metricdepthestimationmonocularlatentdiffusionmodelsStablefine-tuningzero-shotgeneralizationsyntheticRGB-Ddatalog-scaledrepresentation

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 a text-to-image generative model can be converted into a monocular metric depth estimator, a system that predicts a distance in metres for every pixel of a single photograph. MetricGold starts from Stable Diffusion, fine-tunes its variational autoencoder on log-normalized metric depth, and fine-tunes its denoising U-Net to translate an RGB image into a depth latent, using only photorealistic synthetic RGB-D data. The authors report that this protocol trains in about two days on a single consumer GPU and, without ever seeing real depth maps, zero-shot generalizes to real datasets, producing sharper and accurate metric depth estimates than existing approaches. The payoff, if correct, is that the visual knowledge inside a pretrained image generator can substitute for massive collections of real depth measurements.

What carries the argument

The load-bearing mechanism is a fine-tuned latent diffusion pipeline. Stable Diffusion's VAE is re-trained as a Depth VAE to reconstruct log-normalized metric depth maps, while the denoising U-Net is fine-tuned with the RGB latent concatenated to the noisy depth latent, the first-layer weights duplicated and halved so the pretrained structure is preserved. The log-scaled depth representation, $d_\text{log} = \operatorname{normalize}\bigl(\log(d_r/d_{\min}) / \log(d_{\max}/d_{\min})\bigr)$, spreads the depth distribution so that indoor scenes around a few metres and outdoor scenes up to tens of metres share one model; the paper identifies latent depth reconstruction as the main performance bottleneck.

What would settle it

Take a real image with known depth, such as a LiDAR-registered KITTI frame or an indoor RGB-D scan, run the published MetricGold checkpoint, and compare its output in metres with the ground truth using absolute relative error. If the model's predictions are only correct up to an unknown per-image scaling that must be computed from the ground truth min/max depth, or if cropping the image changes the predicted scale, the metric-recovery claim fails. This experiment also discloses whether the inverse of Eq. (3) exists as a deployable transform.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that latent diffusion priors from a text-to-image model are sufficient for monocular metric depth estimation. Depth prediction is reframed as conditional denoising in a compact latent space: an RGB image and a log-normalized metric depth map are each encoded with a variational autoencoder, the two latents are concatenated along the channel dimension, and a fine-tuned U-Net learns to remove noise from corrupted depth latents conditioned on the image. The only metric-depth supervision comes from photorealistic synthetic datasets, Hypersim for indoor scenes and Virtual KITTI 2 for outdoor driving. MetricGold then transfers to real datasets without retraining and outputs depth in metric units, rather than relative or affine-invariant depth, which is the qualitative step beyond prior diffusion-based depth estimators.

Load-bearing premise

The load-bearing premise is that Eq. (3)'s log-normalized depth can be inverted to recover true metres on arbitrary real images, but the paper does not state whether the smallest and largest depth values, $d_{\min}$ and $d_{\max}$, are per-image ground-truth bounds, dataset-global constants, or learned values, nor what inverse transform is applied at test time.

Editorial extensions

If this is right

A consumer-grade fine-tuning recipe—one RTX 3090 and roughly two days—can turn a pretrained text-to-image diffusion model into a zero-shot metric depth estimator, making such models practical for academic labs.
Training exclusively on photorealistic synthetic data avoids sensor-specific noise and bias that real RGB-D datasets carry, which the paper argues yields sharper depth boundaries.
Since MetricGold operates in latent space with v-parameterization and 50-step DDIM sampling, the same model is a candidate for consistency distillation, bringing inference to a handful of steps.

Reading between the lines

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

The paper leaves the inverse of Eq. (3) unspecified; a concrete next step is to report the exact $d_{\min}$, $d_{\max}$, and un-normalization used at inference, because without that the metric claim is not independently checkable.
If the recipe generalizes, it suggests a broader pattern: any strong latent diffusion model can be repurposed for dense visual prediction by fine-tuning only the output VAE and the U-Net's input layer, which could apply to surface normals, segmentation, or optical flow.
Because the metric scale is learned entirely from synthetic renderers, its accuracy on camera systems with unusual intrinsics or extreme depth ranges, such as fisheye or long-range LiDAR-only setups, is an untested boundary that would show where the diffusion prior stops carrying the scale.

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.

Desk Editor's Note

read the letter

Colleague,

Quick take: MetricGold is a plausible fine-tuning recipe — adapt Stable Diffusion's latent space to output log-scaled metric depth, train on synthetic RGB-D only — but the preprint currently contains no experimental evidence. The central claim of zero-shot metric generalization is asserted, not shown. The stress-test concern is right: Eq. (3) defines dlog without ever saying where dmin/dmax come from or how the decoder output is inverted back to metric units. Without that, the method isn't reproducible, and the "metric" part of the contribution is literally undefined.

What is actually new: the combination of latent-space diffusion (following Marigold) with DMD's metric-depth objective and Depth Anything V2's synthetic-only training is a reasonable next step, and the authors correctly identify DMD as the closest prior work. The training details — concatenating image and depth latents, doubling the U-Net input weights and halving them, fine-tuning a Depth VAE on log-normalized depth — are described concretely enough that someone could try to reproduce the pipeline. Training time on one RTX 3090 is a nice practical claim, and two days is plausible.

The soft spots are not minor. There is no experiments section: no tables, no error bars, no comparisons. The abstract and conclusion say the method produces "sharper and accurate metric depth estimates compared to existing approaches" but no measured support appears anywhere in the manuscript. This is the load-bearing claim, and it is unverified. Second, the normalization/inverse gap is technical and load-bearing: if dmin/dmax are per-image ground-truth bounds, inference requires knowing the answer; if they are dataset constants, they need to be stated, and the inverse transform must be exact. The paper reports neither. The linked code might resolve some of this, but I did not audit it, and the manuscript alone does not stand.

I'd rather not cite this in its current form, and I would not send it to peer review as-is. It could become a short technical report with real numbers and a clear definition of the inverse transform. As it stands, a serious referee would bounce it for lack of evidence, and I would too. If you are working on diffusion-based depth, the recipe is worth a skim, but do not treat the claims as established.

Referee Report

3 major / 5 minor

Summary. The paper proposes MetricGold, a fine-tuning protocol that repurposes Stable Diffusion v2's latent diffusion model for monocular metric depth estimation. The method replaces the Stable Diffusion VAE with a depth VAE trained to reconstruct log-normalized metric depth from synthetic RGB-D data (Hypersim, Virtual KITTI, and TartanAir are named), concatenates image and depth latents as conditioning for the denoising U-Net, and fine-tunes the U-Net with a standard diffusion objective. The authors claim that this protocol trains efficiently on a single RTX 3090 in about two days and yields zero-shot generalization to unseen real datasets with sharper and more accurate metric depth estimates than existing approaches. The manuscript contains no quantitative evaluation, no benchmark tables, and no explicit inverse for the log-depth normalization.

Significance. If the central claims were substantiated, MetricGold would be a useful demonstration that a consumer-grade fine-tuning recipe can convert a text-to-image diffusion model into a zero-shot metric depth estimator using only synthetic data. The paper has some genuine strengths: it releases code, reports a concrete training budget, builds on external baselines rather than circular self-citation, and identifies a plausible design space (latent-space diffusion, log-scaled depth, synthetic-only training). However, the absence of any experimental validation makes the significance currently unverified; the claimed advantages over Marigold, DMD, and Depth Anything V2 are not established, and the metric-scale recovery mechanism is not fully specified.

major comments (3)

[Abstract; Section 5] The paper's central claim of "robust generalization" and "sharper and accurate metric depth estimates compared to existing approaches" is not supported by any quantitative evaluation. There is no experiments section, no benchmark tables, no error metrics (such as AbsRel, RMSE, or delta thresholds), no comparisons to Marigold, DMD, Depth Anything V2, or any other baseline, and no qualitative result figures beyond the teaser. The conclusion in Section 5 repeats the empirical claim without presenting evidence. This is load-bearing because the abstract explicitly promises experimental demonstration, and it makes the paper's central contribution untestable as written.
[Section 3.2, Eq. (3)] The log-depth normalization is underspecified in a way that affects the core claim of metric depth recovery. Eq. (3) defines dlog = normalize(log(dr/dmin)/log(dmax/dmin)), but the paper does not state whether dmin and dmax are per-image ground-truth bounds, dataset-global constants, or learned values, and it does not give the inverse transform used to recover dr from the network output at inference. If per-image bounds are used, the output cannot be rescaled to metric units at inference without knowing the answer in advance; if global constants are used, they must be reported and the inverse must be exact. Since metric scale recovery is the distinguishing contribution relative to relative-depth models such as Marigold, this omission makes the method irreproducible and the zero-shot metric-depth claim unsupported.
[Section 3.1; Section 4] There is an unresolved mismatch between the diffusion formulation and the reported training objective. Section 3.1 presents a standard noise-prediction objective, while Section 4 states that the model is trained with v-parameterization, which uses a different target. The paper does not reconcile these two descriptions or state the actual training loss used. This is a secondary but still load-bearing reproducibility issue because the reader cannot determine what objective was optimized or which inference procedure corresponds to the reported claims.

minor comments (5)

[Section 3.2] There are typographical errors: "doamins" should be "domains" and "WIth minimal changes" should be "With minimal changes."
[Abstract; Section 1] The name "MariGold" in the abstract and Section 1 should be "Marigold" to match the cited work [6].
[Abstract; Section 2.2; Section 4] The training data are described inconsistently: the abstract lists HyperSIM, VirtualKitti, and TartanAir, while Section 2.2 names only Hypersim and Virtual KITTI 2, and Section 4 gives no dataset composition. Please clarify which datasets were actually used and in what proportions.
[Section 3.2, Eq. (4)] The VAE loss in Eq. (4) is written as L_VAE = E_{q(d|x)}[log p(log dlog | d_latent)] - D_KL(q(d_latent|x) || p(d_latent)), but the notation is confusing: dlog is described as a deterministic input to the encoder, not as a random variable conditioned on x. Please clarify the generative model and the exact reconstruction loss used for the depth VAE.
[References] Reference [4] has a malformed author list ("Rare s, Ambrus, , and Adrien Gaidon"), and the paper uses "DMD" without spelling out the method name or giving a dedicated reference; [15] appears to be the DMD paper but should be cited explicitly at first use.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: MetricGold is an empirical fine-tuning recipe built on external baselines and synthetic RGB-D data; no prediction reduces to its own input by construction.

full rationale

MetricGold's load-bearing steps are: (i) fine-tuning Stable Diffusion's VAE and U-Net on synthetic RGB-D pairs to model log-normalized metric depth, and (ii) evaluating zero-shot generalization on unseen real datasets. Neither step reduces to its own input. The diffusion objective and the VAE loss in Eq. (4) train on paired (x, d) data, and the decoder output is presented as the predicted depth map. The paper explicitly derives its initialization from external work (Marigold [6], DDVM [15], Depth Anything V2 [17], Stable Diffusion [13]) and does not invoke any self-citation as a load-bearing premise. There is no fitted parameter that is later renamed as a prediction, and no uniqueness theorem is imported from the authors' prior work. The one concern worth noting is not circular: Eq. (3)'s log-depth normalization references dmin and dmax without specifying whether these are per-image ground-truth bounds or dataset-global constants, and the inverse transform back to metric depth is never stated. If per-image bounds were used, the metric-scale claim would be vacuous because the target already encodes the answer; if global constants were used, they would need to be reported. However, the text gives no basis to conclude either usage, so this is an underspecification/reproducibility and evidential gap, not a demonstrated circular reduction. Under the hard rule that circularity requires quoting a specific reduction, no step qualifies.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central claim rests on two synthetic datasets, one pretrained text-to-image model, and an underspecified log-depth normalization. No new physical entities are introduced.

free parameters (4)

log-depth normalization bounds dmin/dmax = not reported
Eq. (3) normalizes depth with these bounds; the paper never reports their values or whether they are per-image, per-dataset, or global. Metric scale recovery at inference depends on them.
input layer weight scaling factor = 0.5
Section 3.2 halves the duplicated first-layer weights to preserve activation magnitudes; a hand-chosen constant that affects initialization.
inference aggregation count = 10
Section 4 averages 10 DDIM runs with varying noise; chosen without ablating the trade-off between compute and accuracy.
DDIM inference steps = 50
Section 4 switches from 1000 DDPM training steps to 50 DDIM steps; no validation is reported for this choice.

assumptions (4)

domain assumption Stable Diffusion's pretrained latent space contains visual priors transferable to metric depth regression.
Motivates the whole approach in Section 1 and Section 3.2; no ablation isolates the contribution of the diffusion prior.
domain assumption Training only on Hypersim and Virtual KITTI 2 suffices for zero-shot indoor and outdoor metric depth.
Section 2.2 selects these two synthetic datasets; no coverage or domain-gap analysis is given.
ad hoc to paper Log-normalized depth can be decoded to absolute metric depth without knowing per-image dmin and dmax.
Eq. (3) defines log depth with dmin/dmax; the inverse mapping for unseen images is unspecified, making this assumption paper-specific.
domain assumption The DDIM 50-step sampler with 10-run averaging faithfully represents the learned depth distribution.
Section 4 fixes these inference settings without comparing to full DDPM sampling.

how reviews work

0 comments

Cite this review

Pith. "Pith review of MetricGold: Leveraging Text-To-Image Latent Diffusion Models for Metric Depth Estimation." pith.science (2026). https://pith.science/paper/CZTWRNXH

@misc{pith2026241110886,
  author       = {Pith},
  title        = {Pith review of: MetricGold: Leveraging Text-To-Image Latent Diffusion Models for Metric Depth Estimation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CZTWRNXH}},
  note         = {Machine review of arXiv:2411.10886}
}

read the original abstract

Recovering metric depth from a single image remains a fundamental challenge in computer vision, requiring both scene understanding and accurate scaling. While deep learning has advanced monocular depth estimation, current models often struggle with unfamiliar scenes and layouts, particularly in zero-shot scenarios and when predicting scale-ergodic metric depth. We present MetricGold, a novel approach that harnesses generative diffusion model's rich priors to improve metric depth estimation. Building upon recent advances in MariGold, DDVM and Depth Anything V2 respectively, our method combines latent diffusion, log-scaled metric depth representation, and synthetic data training. MetricGold achieves efficient training on a single RTX 3090 within two days using photo-realistic synthetic data from HyperSIM, VirtualKitti, and TartanAir. Our experiments demonstrate robust generalization across diverse datasets, producing sharper and higher quality metric depth estimates compared to existing approaches.

Figures

Figures reproduced from arXiv: 2411.10886 by the authors.

**Figure 2.** Overview of the MetricGold fine-tuning protocol: Beginning with a pretrained Stable Diffu [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

19 extracted references · 12 canonical work pages

[1]

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, Wesam Manassra, Prafulla Dhariwal, Casey Chu, Yunxin Jiao, and Aditya Ramesh. Improving image generation with better captions. https: // cdn. openai. com/ papers/ dall-e-3. pdf , 2023

work page 2023
[2]

Virtual KITTI 2

Yohann Cabon, Naila Murray, and Martin Humenberger. Virtual KITTI 2. arXiv preprint arXiv:2001.10773, 2020. 5

arXiv 2001
[3]

Omnidata: A scalable pipeline for making multi-task mid-level vision datasets from 3d scans

Ainaz Eftekhar, Alexander Sax, Jitendra Malik, and Amir Zamir. Omnidata: A scalable pipeline for making multi-task mid-level vision datasets from 3d scans. In ICCV, pages 10786–10796, 2021

work page 2021
[4]

Towards zero-shot scale-aware monocular depth estimation

Vitor Guizilini, Igor Vasiljevic, Dian Chen, Rare s, Ambrus, , and Adrien Gaidon. Towards zero-shot scale-aware monocular depth estimation. In ICCV, 2023

work page 2023
[5]

Distilling diffusion models into conditional gans, 2024

Minguk Kang, Richard Zhang, Connelly Barnes, Sylvain Paris, Suha Kwak, Jaesik Park, Eli Shechtman, Jun-Yan Zhu, and Taesung Park. Distilling diffusion models into conditional gans, 2024

work page 2024
[6]

Repurposing diffusion-based image generators for monocular depth estimation, 2024

Bingxin Ke, Anton Obukhov, Shengyu Huang, Nando Metzger, Rodrigo Caye Daudt, and Kon- rad Schindler. Repurposing diffusion-based image generators for monocular depth estimation, 2024

work page 2024
[7]

V A-DepthNet: A variational approach to single image depth prediction

Ce Liu, Suryansh Kumar, Shuhang Gu, Radu Timofte, and Luc Van Gool. V A-DepthNet: A variational approach to single image depth prediction. In ICLR, 2023

work page 2023
[8]

Latent consistency models: Synthesizing high-resolution images with few-step inference, 2023

Simian Luo, Yiqin Tan, Longbo Huang, Jian Li, and Hang Zhao. Latent consistency models: Synthesizing high-resolution images with few-step inference, 2023

2023

Show all 19 references

[9]

Dinov2: Learning robust visual features without supervision, 2024

Maxime Oquab, Timothée Darcet, Théo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, Mahmoud Assran, Nicolas Ballas, Wojciech Galuba, Russell Howes, Po-Yao Huang, Shang-Wen Li, Ishan Misra, Michael Rabb...

2024
[10]

Vision transformers for dense prediction

René Ranftl, Alexey Bochkovskiy, and Vladlen Koltun. Vision transformers for dense prediction. In ICCV, 2021

2021
[11]

Towards robust monocular depth estimation: Mixing datasets for zero-shot cross-dataset transfer

René Ranftl, Katrin Lasinger, David Hafner, Konrad Schindler, and Vladlen Koltun. Towards robust monocular depth estimation: Mixing datasets for zero-shot cross-dataset transfer. IEEE TPAMI, 2020

2020
[12]

Susskind

Mike Roberts, Jason Ramapuram, Anurag Ranjan, Atulit Kumar, Miguel Angel Bautista, Nathan Paczan, Russ Webb, and Joshua M. Susskind. Hypersim: A photorealistic synthetic dataset for holistic indoor scene understanding. In ICCV, 2021

2021
[13]

High- resolution image synthesis with latent diffusion models

Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. High- resolution image synthesis with latent diffusion models. In CVPR, pages 10684–10695, 2022

2022
[14]

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. NeurIPS, 35, 2022

2022
[15]

Saurabh Saxena, Junhwa Hur, Charles Herrmann, Deqing Sun, and David J. Fleet. Zero-shot metric depth with a field-of-view conditioned diffusion model, 2023

2023
[16]

Tartanair: A dataset to push the limits of visual slam, 2020

Wenshan Wang, Delong Zhu, Xiangwei Wang, Yaoyu Hu, Yuheng Qiu, Chen Wang, Yafei Hu, Ashish Kapoor, and Sebastian Scherer. Tartanair: A dataset to push the limits of visual slam, 2020

2020
[17]

Depth anything v2, 2024

Lihe Yang, Bingyi Kang, Zilong Huang, Zhen Zhao, Xiaogang Xu, Jiashi Feng, and Hengshuang Zhao. Depth anything v2, 2024

2024
[18]

Metric3D: Towards zero-shot metric 3d prediction from a single image

Wei Yin, Chi Zhang, Hao Chen, Zhipeng Cai, Gang Yu, Kaixuan Wang, Xiaozhi Chen, and Chunhua Shen. Metric3D: Towards zero-shot metric 3d prediction from a single image. In ICCV, 2023

2023
[19]

NeWCRFs: Neural window fully-connected CRFs for monocular depth estimation

Weihao Yuan, Xiaodong Gu, Zuozhuo Dai, Siyu Zhu, and Ping Tan. NeWCRFs: Neural window fully-connected CRFs for monocular depth estimation. In CVPR, 2022. 6

2022

Pith tools

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

[1] [1]

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, Wesam Manassra, Prafulla Dhariwal, Casey Chu, Yunxin Jiao, and Aditya Ramesh. Improving image generation with better captions. https: // cdn. openai. com/ papers/ dall-e-3. pdf , 2023

work page 2023

[2] [2]

Virtual KITTI 2

Yohann Cabon, Naila Murray, and Martin Humenberger. Virtual KITTI 2. arXiv preprint arXiv:2001.10773, 2020. 5

arXiv 2001

[3] [3]

Omnidata: A scalable pipeline for making multi-task mid-level vision datasets from 3d scans

Ainaz Eftekhar, Alexander Sax, Jitendra Malik, and Amir Zamir. Omnidata: A scalable pipeline for making multi-task mid-level vision datasets from 3d scans. In ICCV, pages 10786–10796, 2021

work page 2021

[4] [4]

Towards zero-shot scale-aware monocular depth estimation

Vitor Guizilini, Igor Vasiljevic, Dian Chen, Rare s, Ambrus, , and Adrien Gaidon. Towards zero-shot scale-aware monocular depth estimation. In ICCV, 2023

work page 2023

[5] [5]

Distilling diffusion models into conditional gans, 2024

Minguk Kang, Richard Zhang, Connelly Barnes, Sylvain Paris, Suha Kwak, Jaesik Park, Eli Shechtman, Jun-Yan Zhu, and Taesung Park. Distilling diffusion models into conditional gans, 2024

work page 2024

[6] [6]

Repurposing diffusion-based image generators for monocular depth estimation, 2024

Bingxin Ke, Anton Obukhov, Shengyu Huang, Nando Metzger, Rodrigo Caye Daudt, and Kon- rad Schindler. Repurposing diffusion-based image generators for monocular depth estimation, 2024

work page 2024

[7] [7]

V A-DepthNet: A variational approach to single image depth prediction

Ce Liu, Suryansh Kumar, Shuhang Gu, Radu Timofte, and Luc Van Gool. V A-DepthNet: A variational approach to single image depth prediction. In ICLR, 2023

work page 2023

[8] [8]

Latent consistency models: Synthesizing high-resolution images with few-step inference, 2023

Simian Luo, Yiqin Tan, Longbo Huang, Jian Li, and Hang Zhao. Latent consistency models: Synthesizing high-resolution images with few-step inference, 2023

2023