Pith. sign in

REVIEW 3 major objections 6 minor 46 references

Localize, Don't Beautify: Client-Side Control of Image-Editing APIs for Cosmetic Surgery Previews

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

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

Pith's one-line read Client-side mask-and-paste confines AI face edits to the intended region without any access to the model.

desk verdict A small, candid pilot whose localization gain is near-tautological; the real empirical claim—that compositing doesn't suppress the intended edit—rests on one sample per cell and needs replication before the benchmark is trusted. read the letter →

arxiv 2608.02841 v1 pith:GTMSC3NP submitted 2026-08-03 cs.CV

Sukhrobbek Ilyosbekov This is my paper · ORCID
classification cs.CV
keywords client-sideimageeditingmaskedcompositingblack-boxAPIsregion-confinedcosmeticsurgerypreviewsidentitypreservationCIELABlocalizationmetricinstruction-guided
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

Commercial image-editing APIs change far more of a face than a prompt names, and existing methods that pin an edit to one region need the model's internals, which these services do not expose. This paper asks how much region control can be achieved from the client side alone and reports that a simple post-processing step delivers it: cut the edited region out of the API response, align it to the original photo, and paste it back through a feathered mask derived from facial landmarks. Across six commercial editing configurations and one mask-based inpainting service, the masked composite raised the measured localization ratio over paired prompt-only outputs by a median of 0.446 on 12 frontal faces while leaving target-region pixel change nearly unchanged (median $\Delta E_{\mathrm{tgt}}$ 4.5 versus 4.0). The same benchmark found that no tested editor moved outputs closer to the face's own postoperative photograph in face-recognition embedding space, and the one tested inpainting model did not beat the client-side composite. The study is explicitly a control diagnostic with one generation per cell and no clinical ratings, but its core point is that model access is not needed to enforce region confinement.

What carries the argument

The load-bearing mechanism is the masked composite, a three-step client-side wrapper. First, a region mask is built from a 478-point face-mesh landmark set: the face splits at the lower-eyelid line into an upper-face keep zone and a jaw--neck target zone, and the nose region is a dilated perimeter hull. Then the edited response is aligned back onto the original photograph with a similarity warp estimated from stable landmarks. Finally it is blended as $\mathrm{out} = x(1-w) + e\,w$, where $x$ is the input, $e$ is the aligned edit, and $w$ is a feathered per-pixel weight; compositing is skipped when estimated yaw exceeds 0.35 to avoid ghost seams on profile views. The diagnostic carrying the quantitative claim is the localization ratio $\mathrm{loc} = \Delta E_{\mathrm{tgt}} / (\Delta E_{\mathrm{tgt}} + \Delta E_{\mathrm{off}})$, a CIELAB pixel-change fraction saying how much measured change lands inside the requested region rather than in the protected upper-face zone.

What would settle it

Re-run the benchmark with repeated generations per cell on the same 12 frontal faces, for example 10 outputs per model and control, and recompute the paired localization gain; the central claim weakens if the face-clustered interval no longer excludes zero or if any editor's composited outputs show off-mask $\Delta E$ comparable to its prompt-only outputs.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that text instructions alone cannot confine a black-box editor to the requested facial region, while a client-side masked composite can. For the six primary editing configurations, composited outputs had a median localization ratio of 0.985 (range 0.968--0.998) against 0.538 (range 0.347--0.680) for the same prompts without compositing; the within-face, within-model median gain was 0.446 with a face-clustered bootstrap interval of 0.421--0.457. Median off-target pixel change $\Delta E_{\mathrm{off}}$ fell from 3.50 to 0.07, while median target-region change $\Delta E_{\mathrm{tgt}}$ stayed at roughly the same level (4.5 versus 4.0), meaning the composite removed leakage without erasing the requested edit. The tested mask-based inpainting model did not surpass this post-hoc control, and against each face's input-to-postoperative baseline all editors showed a negative median identity change of $-0.029$, so none moved closer to the postoperative photograph in embedding space. The paper treats these as system-level measurements of control, not clinical evidence of surgical accuracy.

Load-bearing premise

The load-bearing premise is that one generated output per model-and-photo combination fairly represents how that editor behaves; if the single output is atypical, the ranking of editors, the composite-versus-inpainting comparison, and the claim that target-region change stays level could all shift.

Editorial extensions

If this is right

  • Any editor reachable through a public endpoint can be region-confined without provider cooperation, at the cost of one alignment and one feathering operation per edit.
  • Prompt guardrails alone are not sufficient for surgical previews: prompt-only outputs still changed the protected keep zone, so localization has to be enforced in pixel space.
  • Native mask-based inpainting is not automatically better than client-side compositing; the one inpainting model tested produced a near-copy on rhinoplasty and lower localization than the composite on the same task.
  • Choosing an editor for preview work becomes an edit-strength-versus-identity trade-off rather than a pure quality ranking, since editors span a wide range of target-region change and identity retention.
  • On the embedding-level outcome check, current editors moved outputs slightly away from the postoperative photograph rather than toward it, so these APIs do not yet act as outcome predictors under the protocol tested.

Reading between the lines

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

  • A natural extension is repeated sampling: because the paper generates one output per cell and two editors ignore the supplied seed, running many generations per cell could convert the observed 0.446 median gain into a distribution and reveal editors where compositing matters less.
  • The near-perfect localization of the composite is partly by construction, since off-mask pixels are copied verbatim; the more informative follow-up would measure whether inside-mask edits remain anatomically plausible, for example by comparing landmark geometry before and after.
  • The same wrapper should transfer to other black-box editing tasks beyond faces, such as product or document edits, whenever a mask, an alignment, and a feathered blend can be derived from the image content.
  • The negative identity movement relative to the postoperative baseline may reflect session differences in pose, expression, and lighting rather than anatomical failure; a pose-normalized geometry comparison would separate identity-embedding drift from true outcome drift.
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 studies whether region-confined editing of black-box commercial image-editing APIs can be achieved entirely on the client side, for cosmetic-surgery previews. It proposes a control ladder: prompt-only, a masked composite that aligns the model's output back onto the input and blends it through a landmark-derived feathered mask, and native masked inpainting where supported. On a pilot benchmark of six commercial editing configurations and one inpainting model, the paper reports that masked compositing increases a CIELAB localization ratio by a median of 0.446 on 12 frontal faces while leaving target-region pixel change roughly unchanged (median ΔE_tgt 4.5 vs 4.0), that the tested inpainting model did not beat the composite, and that no editor moved ArcFace embeddings closer to postoperative photographs. The authors explicitly scope the study as a pilot, disclaim clinical validity, and note that each cell contains one generated output and that two models ignore the supplied seed.

Significance. If the descriptive claims hold, the paper makes a useful systems-level observation: for black-box editing APIs, off-target pixel change can be suppressed entirely on the client, at trivial cost and without model internals. The control ladder is a simple, reproducible baseline, and the paper is transparent about its limitations, provides code and a canonical score table, and avoids overclaiming clinical relevance. The main non-tautological empirical claim—that compositing preserves on-target pixel change—is not, however, supported by the single-sample design. The localization gain itself is a mathematical consequence of the compositing construction, so the paper's significance rests on the target-region and editor-difference results, which are exactly the parts threatened by generation stochasticity. The paper is an honest pilot, but its abstract and conclusion currently state the target-preservation result more strongly than the evidence allows.

major comments (3)
  1. [Sec. V and Sec. VI-A] The claim in Sec. VI-A that masked compositing retains target-region pixel change (median ΔE_tgt 4.5 vs 4.0) is not supported by the experimental design, because Sec. V states that one output is generated per cell and Sec. VII confirms that two models ignore the supplied seed. The face-clustered bootstrap interval for the localization gain does not account for generation stochasticity, so a single atypical draw could flip or widen the 0.5-unit difference. The authors should either add repeated generations (e.g., 5–10 per cell) and report within-model variance, or explicitly downgrade this claim to a preliminary observation that requires further testing.
  2. [Sec. IV-B, Eq. (1)] The localization ratio loc is forced toward 1 for the masked composite because compositing copies off-mask pixels from the input by construction, making ΔE_off approximately zero. The paper acknowledges this in Sec. IV-B and Sec. VII, but the abstract and conclusion present the 0.446 gain as a primary result. Since the gain is a mathematical consequence of the construction rather than an empirical property of the editors, the non-tautological content resides in the target-region and editor-difference results, exactly the parts threatened by single-sample stochasticity. The contribution should be reframed so the localization gain is reported as a consistency check of the compositing implementation, not as a headline improvement.
  3. [Sec. VI-C] The comparison between Qwen-Image-Edit's masked inpainting and client-side compositing rests on a single model, a single output per cell, and shows the inpainter producing near-copies on rhinoplasty (ΔE_tgt 1.9, identity 0.965). This near-copy behavior suggests the model may not have executed the requested edit, so the conclusion that 'client-side compositing produced stronger measured keep-zone preservation than provider-side masking' is confounded by possible instruction-following failure. The authors should either add repeated generations, test additional inpainting models, or report qualitative evidence that the requested anatomical change was actually made, to allow a fair comparison.
minor comments (6)
  1. [Abstract] The abstract states that the masked composite improved localization 'while changing the requested region about as much,' but the single-sample design and the tautological nature of the localization metric (Sec. IV-B) make this statement stronger than the evidence supports; the abstract should carry the same caveats as Sec. VII.
  2. [Sec. III-B] The definition of the keep zone for rhinoplasty deserves clarification: the text says the keep zone is 'the upper-face region minus the target,' but the nose region is defined by a hull that likely extends below the lower-eyelid line, so it is not obvious that subtracting the target from the upper-face zone yields a meaningful disjoint keep zone. Please spell out how the keep zone is computed for the nose procedure.
  3. [Table II] The column header 'Lat.' is ambiguous; it presumably denotes latency in seconds, but the abbreviation is unusual. Spell it out or use 'Latency (s)'.
  4. [Sec. V] The description of the face set is difficult to follow: 'The benchmark matrix uses 8 facelift faces and 8 rhinoplasty faces (the 4 frontal cross-views, which the regional metric can score, plus 4 profiles...)' – it would help to state explicitly that the main analysis contains 8 facelift and 7 rhinoplasty faces after one extreme-profile face failed detection.
  5. [Sec. VI-D] Reporting 'only 15% of outputs score above their face's baseline' as a percentage is unhelpful without the denominator; report the count (e.g., 29 of 194) alongside the percentage.
  6. [Sec. IX] The conclusion's phrase 'uniformly across the six editors tested' refers to off-target preservation, which is a by-product of the compositing construction; consider rephrasing to avoid implying that the editors themselves behaved uniformly.

Circularity Check

2 steps flagged · score 6.0 of 10

Composite localization gain is a definitional artifact, as the paper's own limitation text concedes.

  1. self definitional [Sec. IV-B (third listed limitation); reported as a result in Sec. VI-A; blend definition in Sec. III-C.]
    "Third, compositing copies off-mask pixels by design, so a high ratio on that rung verifies the implementation’s preservation property rather than the editor’s quality."

    loc is defined as ΔE_tgt/(ΔE_tgt+ΔE_off), and the composite is defined as out = x(1−w)+e w with w=0 outside the mask, so off-mask output is pixel-identical to the input and ΔE_off is forced to ~0. Therefore loc→1 and the reported median gain of 0.446 over prompt-only (Sec. VI-A) is an algebraic consequence of the construction, not a measured property of the editors. The paper itself concedes this in the quoted sentence and again in Sec. VI-A ('Post-hoc compositing did so because preservation is enforced by the blend'). The headline claim that the composite 'enforces off-mask preservation' restates the definition of the composite rather than testing a model behavior.

  2. self definitional [Sec. VI-C, comparison of masked inpaint vs. masked composite.]
    "For this model and these single outputs, client-side compositing produced stronger measured keep-zone preservation than provider-side masking."

    The keep-zone preservation being compared is the same construction-forced property: with w=0 outside the blend mask, composite ΔE_off≈0 and loc≈1 by definition, whereas Qwen's inpaint output is a genuine model output with measured off-target change. The conclusion that the composite 'did not beat' or outperformed the inpainter on localization is therefore pre-ordained by the metric-construction pairing, not by editor behavior. The paper's restriction to 'measured keep-zone preservation' and to one model narrows the claim but does not remove the conflation of a defined zero with an empirical outcome.

full rationale

The paper's most prominent quantitative result, the median 0.446 localization gain from masked compositing, is not an empirical finding about the editing APIs: the composite is defined as out = x(1−w)+e w with w=0 outside the mask, so ΔE_off is forced to zero and loc toward one. The paper explicitly flags this in Sec. IV-B and Sec. VI-A, yet still reports the gain as a headline contribution and derives the 'across every editor tested' claim from it. The same definitional mechanism drives the composite-versus-inpaint comparison on the keep-zone axis. Genuinely non-circular content remains: the paired ΔE_tgt values (median 4.5 composite vs 4.0 prompt-only), the ArcFace identity distributions, and the postoperative-embedding baseline analysis are measured outputs rather than consequences of the construction, so the circularity is partial rather than total. The single-output-per-cell design and the two seed-ignoring models are statistical weaknesses, not circularity, and no load-bearing self-citation chain is present; the Envisage citation is external acknowledgment of the identity-confounding effect. Because the headline localization contribution reduces by construction while independent empirical content survives, the score is 6 rather than 8 or 10.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

No new entities or fitted statistical parameters appear. The mask constants are hand-chosen geometry, and the diagnostic axioms are borrowed from prior work rather than derived, so the reader supplies the validity of ArcFace, CIELAB, and the landmark masks.

free parameters (3)
  • jaw-neck hull extension below chin = fixed fraction of face height (value not stated)
    Hand-chosen mask geometry in Sec. III-B; defines how far the jaw-neck composite extends below the chin and therefore what the metric counts as target change.
  • nose region dilation margin = proportional to inter-ocular distance (multiplier not stated)
    Hand-chosen per Sec. III-B; scales the nose mask, feather, and erosion radii, affecting both composite seams and the target/keep-zone partition.
  • pose gate yaw threshold = 0.35
    Hand-selected threshold in Sec. III-C; determines which profile outputs are left uncomposited, changing which cells contribute to composite localization summaries.
assumptions (4)
  • domain assumption ArcFace cosine similarity is a valid proxy for identity preservation and for movement toward a postoperative photograph.
    Sec. IV-A and IV-C use ArcFace for identity and post-op reference; the 0.6 line is a heuristic and embedding direction is not clinically validated.
  • domain assumption Mean CIELAB Delta E76 inside landmark-defined regions is a meaningful measure of on-target versus off-target edit change.
    Sec. IV-B defines loc from Delta E_tgt and Delta E_off; the keep zone is partial, abstains on profiles, and the metric is not perceptual or anatomical.
  • domain assumption MediaPipe face-mesh landmarks correctly delineate forehead, eyes, jaw, chin, and nose regions on frontal faces.
    Sec. III-B builds every mask from the 478-point mesh; the mesh has no neck landmarks, so the jaw-neck hull is extended by a hand-chosen fraction, and profile views require a fallback or gating.
  • domain assumption One output per model-cell is representative of that editor's behavior over the study period.
    Sec. V and VII state a single output per cell, seeds are ignored by some models, and endpoints are moving targets; all measured editor differences depend on this.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Localize, Don't Beautify: Client-Side Control of Image-Editing APIs for Cosmetic Surgery Previews." pith.science (2026). https://pith.science/paper/GTMSC3NP

@misc{pith2026260802841,
  author       = {Pith},
  title        = {Pith review of: Localize, Don't Beautify: Client-Side Control of Image-Editing APIs for Cosmetic Surgery Previews},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GTMSC3NP}},
  note         = {Machine review of arXiv:2608.02841}
}
read the original abstract

Ask a commercial image editor to preview a cosmetic procedure and it will often change more of the face than the request names: a nose edit can also smooth skin or alter lighting. Existing methods for confining an edit to one region require access to the model's internals, which a public editing API does not expose. We ask how much control is possible from the client side alone. In a pilot benchmark, six commercial editing configurations and one mask-based inpainting model perform facelift-style jaw-neck and rhinoplasty edits at three levels of client-side control: the prompt alone; cutting the edited region out of the response and pasting it back onto the original photograph through a landmark-derived mask (a masked composite); and asking the model itself to inpaint inside the mask where supported. Of 210 attempted edits, 196 could be scored. ArcFace cosine measures identity preservation; a CIELAB pixel-change ratio measures how much change lands inside the requested region rather than a protected facial zone. On the 12 frontal faces the regional metric could score, the masked composite improved localization over the paired prompt-only output by a median of 0.446 (95% face-clustered bootstrap interval 0.421-0.457) while changing the requested region about as much. Editors differed in edit strength versus identity retention, and the one inpainting model we tested did not beat the simple composite. Against each face's input-to-postoperative baseline, no editor moved its outputs closer to the postoperative photograph in identity-embedding terms. This is a study of control, not clinical accuracy: no surgeons rated the outputs, and each condition was generated once. Within that scope, keeping a surgical preview inside its intended region needs no access to the model; a mask and composite on the client enforce it across every editor tested, at low provider cost.

Figures

Figures reproduced from arXiv: 2608.02841 by the authors.

Figure 1
Figure 1. A facelift-style jaw–neck edit from one photograph (GPT Image 2). The prompt-only output (center) changes regions throughout the image; the [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Every scoreable output placed by the two automated diagnostics. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. One facelift-style jaw–neck edit composited from every primary editor. The visible edit-strength spectrum parallels the automated identity scores; the [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: (a) Localization by control, models pooled. (b) Identity by model, controls pooled; the dashed line is the heuristic 0.6 reference. Each point is one [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: ArcFace cosine between each edit and the face’s postoperative [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 7
Figure 7. Figure 7: Selected outputs (top to bottom: two facelift-style faces and one frontal rhinoplasty face). Prompt-only outputs alter regions outside the request; in [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 27 canonical work pages

  1. [1]

    Computer imaging and patient satisfaction in rhinoplasty surgery,

    H. R. Sharp, R. S. Tingay, S. Coman, V . Mills, and D. N. Roberts, “Computer imaging and patient satisfaction in rhinoplasty surgery,”The Journal of Laryngology & Otology, vol. 116, no. 12, pp. 1009–1013, 2002

    work page 2002

  2. [2]

    Three-dimensional simulation accuracy and patient satisfaction with rhinoplasty,

    K. Yamamichi, Y . Nakanishi, and C. Y . Chen, “Three-dimensional simulation accuracy and patient satisfaction with rhinoplasty,”Aesthetic Surgery Journal Open Forum, vol. 7, p. ojaf110, 2025

    work page 2025

  3. [3]

    Realistic expectations: To morph or not to morph?

    A. Agarwal, E. Gracely, and W. E. Silver, “Realistic expectations: To morph or not to morph?”Plastic and Reconstructive Surgery, vol. 119, no. 4, pp. 1343–1351, 2007

    work page 2007

  4. [4]

    Blended diffusion for text- driven editing of natural images,

    O. Avrahami, D. Lischinski, and O. Fried, “Blended diffusion for text- driven editing of natural images,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, arXiv:2111.14818

    arXiv 2022

  5. [5]

    DiffEdit: Diffusion-based semantic image editing with mask guidance,

    G. Couairon, J. Verbeek, H. Schwenk, and M. Cord, “DiffEdit: Diffusion-based semantic image editing with mask guidance,” inIn- ternational Conference on Learning Representations (ICLR), 2023, arXiv:2210.11427

    arXiv 2023

  6. [6]

    Turn your vision into reality—AI-powered pre-operative outcome simulation in rhinoplasty surgery,

    S. Knoedler, M. Alfertshofer, S. Simonet al., “Turn your vision into reality—AI-powered pre-operative outcome simulation in rhinoplasty surgery,”Aesthetic Plastic Surgery, vol. 48, pp. 4833–4838, 2024

    work page 2024

  7. [7]

    Envisage: Diffusion-based rhinoplasty goal visualization with mask-decomposed evaluation,

    M. Agarwal and A. D. Bhrany, “Envisage: Diffusion-based rhinoplasty goal visualization with mask-decomposed evaluation,” 2026, arXiv preprint, submitted June 26, 2026

    work page 2026

  8. [8]

    GIE-Bench: Towards grounded evaluation for text-guided image editing,

    Y . Qian, J. Lu, T.-J. Fu, X. Wang, C. Chen, Y . Yang, W. Hu, and Z. Gan, “GIE-Bench: Towards grounded evaluation for text-guided image editing,” 2025

    work page 2025

Show all 46 references
  1. [9]

    GPT-ImgEval: A comprehensive benchmark for diagnosing GPT4o in image generation,

    Z. Yan, J. Ye, W. Li, Z. Huang, S. Yuan, X. He, K. Lin, J. He, C. He, and L. Yuan, “GPT-ImgEval: A comprehensive benchmark for diagnosing GPT4o in image generation,” 2025

    2025

  2. [10]

    The impact of three-dimensional simulation and virtual reality technologies on surgical decision-making and postoperative satisfaction in aesthetic surgery: a preliminary study,

    M. Ersan, H. Demirbas ¸o˘glu, B. Kolcu, and H. A. Yıldız, “The impact of three-dimensional simulation and virtual reality technologies on surgical decision-making and postoperative satisfaction in aesthetic surgery: a preliminary study,”Computer Assisted Surgery, vol. 31, no. ...

    2026

  3. [11]

    A fully automatic postoperative appearance prediction system for blepharoptosis surgery with image-based deep learning,

    Y . Sun, X. Huang, Q. Zhang, S. Y . Lee, Y . Wang, K. Jin, L. Lou, and J. Ye, “A fully automatic postoperative appearance prediction system for blepharoptosis surgery with image-based deep learning,”Ophthalmology Science, vol. 2, no. 3, p. 100169, 2022

    2022

  4. [12]

    PtosisDiffusion: a training-free workflow for precisely predicting post-operative appearance in blepharoptosis patients based on diffusion models,

    S. Huang, J. Xie, B. Yang, Q. Gao, and J. Ye, “PtosisDiffusion: a training-free workflow for precisely predicting post-operative appearance in blepharoptosis patients based on diffusion models,”Frontiers in Cell and Developmental Biology, vol. 12, p. 1459336, 2024

    2024

  5. [13]

    Deep learning-based facial appearance simulation driven by surgically planned craniomaxillofacial bony movement,

    X. Fanget al., “Deep learning-based facial appearance simulation driven by surgically planned craniomaxillofacial bony movement,” inMedical Image Computing and Computer Assisted Intervention (MICCAI), 2022, arXiv:2210.01685

    2022 arXiv

  6. [14]

    Role of artificial intelligence and machine learning in facial aesthetic surgery: A systematic review,

    B. Stephanian, S. Karki, K. Debnath, M. Saltychev, M. Rossi-Meyer, C. K. Kandathil, and S. P. Most, “Role of artificial intelligence and machine learning in facial aesthetic surgery: A systematic review,”Facial Plastic Surgery & Aesthetic Medicine, 2024

    2024

  7. [15]

    The unreasonable effectiveness of deep features as a perceptual metric,

    R. Zhang, P. Isola, A. A. Efros, E. Shechtman, and O. Wang, “The unreasonable effectiveness of deep features as a perceptual metric,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 586–595, arXiv:1801.03924

    2018 arXiv

  8. [16]

    Using generative artificial intelligence tools in cosmetic surgery: A study on rhinoplasty, facelifts, and blepharoplasty proce- dures,

    B. Lim, I. Seth, S. Kah, F. Sofiadellis, R. J. Ross, W. M. Rozen, and R. Cuomo, “Using generative artificial intelligence tools in cosmetic surgery: A study on rhinoplasty, facelifts, and blepharoplasty proce- dures,”Journal of Clinical Medicine, vol. 12, no. 20, p. 6524, 2023

    2023

  9. [17]

    High- resolution image synthesis with latent diffusion models,

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, “High- resolution image synthesis with latent diffusion models,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 10 684–10 695, arXiv:2112.10752

    2022 arXiv

  10. [18]

    SDEdit: Guided image synthesis and editing with stochastic differential equations,

    C. Meng, Y . He, Y . Song, J. Song, J. Wu, J.-Y . Zhu, and S. Ermon, “SDEdit: Guided image synthesis and editing with stochastic differential equations,” inInternational Conference on Learning Representations (ICLR), 2022, arXiv:2108.01073

    2022 arXiv

  11. [19]

    RePaint: Inpainting using denoising diffusion probabilistic models,

    A. Lugmayr, M. Danelljan, A. Romero, F. Yu, R. Timofte, and L. Van Gool, “RePaint: Inpainting using denoising diffusion probabilistic models,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, arXiv:2201.09865

    2022 arXiv

  12. [20]

    Blended latent diffusion,

    O. Avrahami, O. Fried, and D. Lischinski, “Blended latent diffusion,” ACM Transactions on Graphics, vol. 42, no. 4, 2023, sIGGRAPH 2023; arXiv:2206.02779

    2023 arXiv

  13. [21]

    Prompt-to-prompt image editing with cross-attention con- trol,

    A. Hertz, R. Mokady, J. Tenenbaum, K. Aberman, Y . Pritch, and D. Cohen-Or, “Prompt-to-prompt image editing with cross-attention con- trol,” inInternational Conference on Learning Representations (ICLR), 2023, arXiv:2208.01626

    2023 arXiv

  14. [22]

    Plug-and-play diffu- sion features for text-driven image-to-image translation,

    N. Tumanyan, M. Geyer, S. Bagon, and T. Dekel, “Plug-and-play diffu- sion features for text-driven image-to-image translation,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023, arXiv:2211.12572

    2023 arXiv

  15. [23]

    InstructPix2Pix: Learning to follow image editing instructions,

    T. Brooks, A. Holynski, and A. A. Efros, “InstructPix2Pix: Learning to follow image editing instructions,” inIEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), 2023, arXiv:2211.09800

    2023 arXiv

  16. [24]

    MagicBrush: A manually annotated dataset for instruction-guided image editing,

    K. Zhang, L. Mo, W. Chen, H. Sun, and Y . Su, “MagicBrush: A manually annotated dataset for instruction-guided image editing,” in Advances in Neural Information Processing Systems (NeurIPS) Datasets and Benchmarks, 2023, arXiv:2306.10012

    2023 arXiv

  17. [25]

    Imagen editor and EditBench: Advancing and evaluating text-guided image inpainting,

    S. Wang, C. Saharia, C. Montgomeryet al., “Imagen editor and EditBench: Advancing and evaluating text-guided image inpainting,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023, arXiv:2212.06909

    2023 arXiv

  18. [26]

    Emu edit: Precise image editing via recognition and generation tasks,

    S. Sheynin, A. Polyak, U. Singeret al., “Emu edit: Precise image editing via recognition and generation tasks,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 8871– 8879, arXiv:2311.10089

    2024 arXiv

  19. [27]

    ImagenHub: Standardizing the evaluation of conditional image generation models,

    M. Ku, T. Li, K. Zhanget al., “ImagenHub: Standardizing the evaluation of conditional image generation models,” inInternational Conference on Learning Representations (ICLR), 2024, arXiv:2310.01596

    2024 arXiv

  20. [28]

    I2EBench: A comprehensive benchmark for instruction- based image editing,

    Y . Maet al., “I2EBench: A comprehensive benchmark for instruction- based image editing,” inAdvances in Neural Information Processing Systems (NeurIPS), 2024, arXiv:2408.14180

    2024 arXiv

  21. [29]

    HQ-Edit: A high-quality dataset for instruction-based image editing,

    M. Huiet al., “HQ-Edit: A high-quality dataset for instruction-based image editing,” inInternational Conference on Learning Representations (ICLR), 2025, arXiv:2404.09990

    2025 arXiv

  22. [30]

    Step1X-Edit: A practical framework for general image editing,

    S. Liu, Y . Han, P. Xinget al., “Step1X-Edit: A practical framework for general image editing,” 2025, introduces GEdit-Bench

    2025

  23. [31]

    Addressing text embedding leakage in diffusion-based image editing,

    S. Mun, J. Nam, S. Cho, and J. Ok, “Addressing text embedding leakage in diffusion-based image editing,” inIEEE/CVF International Conference on Computer Vision (ICCV), 2025, arXiv:2412.04715; introduces the ALE-Bench attribute-leakage benchmark with target- external/internal l...

    2025 arXiv

  24. [32]

    ArcFace: Additive angular margin loss for deep face recognition,

    J. Deng, J. Guo, N. Xue, and S. Zafeiriou, “ArcFace: Additive angular margin loss for deep face recognition,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp. 4690– 4699, arXiv:1801.07698

    2019 arXiv

  25. [33]

    MediaPipe: A framework for building perception pipelines,

    C. Lugaresi, J. Tang, H. Nashet al., “MediaPipe: A framework for building perception pipelines,” 2019

    2019

  26. [34]

    Attention mesh: High-fidelity face mesh prediction in real-time,

    I. Grishchenko, A. Ablavatski, Y . Kartynnik, K. Raveendran, and M. Grundmann, “Attention mesh: High-fidelity face mesh prediction in real-time,” 2020, cVPR Workshop on Computer Vision for AR/VR

    2020

  27. [35]

    Sample and computation redistribution for efficient face detection,

    J. Guo, J. Deng, A. Lattas, and S. Zafeiriou, “Sample and computation redistribution for efficient face detection,” inInternational Conference on Learning Representations (ICLR), 2022, arXiv:2105.04714

    2022 arXiv

  28. [36]

    The CIEDE2000 color-difference formula: Implementation notes, supplementary test data, and mathemat- ical observations,

    G. Sharma, W. Wu, and E. N. Dalal, “The CIEDE2000 color-difference formula: Implementation notes, supplementary test data, and mathemat- ical observations,”Color Research & Application, vol. 30, no. 1, pp. 21–30, 2005

    2005

  29. [37]

    GPT Image 2,

    OpenAI, “GPT Image 2,” https://developers.openai.com/api/docs/ models/gpt-image-2, 2026, aPI model documentation; released April 21, 2026

    2026

  30. [38]

    Nano Banana Pro: Gemini 3 Pro Image model,

    Google DeepMind, “Nano Banana Pro: Gemini 3 Pro Image model,” https://blog.google/technology/ai/nano-banana-pro/, 2025, announced November 20, 2025

    2025

  31. [39]

    Build with Nano Banana 2, our best image generation and editing model,

    ——, “Build with Nano Banana 2, our best image generation and editing model,” https://blog.google/innovation-and-ai/technology/ developers-tools/build-with-nano-banana-2/, 2026, gemini 3.1 Flash Im- age; announced February 26, 2026

    2026

  32. [40]

    Deeper thinking, more accurate generation: Introduc- ing Seedream 5.0 Lite,

    ByteDance Seed, “Deeper thinking, more accurate generation: Introduc- ing Seedream 5.0 Lite,” ByteDance Seed model announcement, 2026, blog announcement, February 13, 2026; no technical report at time of writing

    2026

  33. [41]

    Seedream 4.0: Toward next-generation multimodal image gener- ation,

    ——, “Seedream 4.0: Toward next-generation multimodal image gener- ation,” 2025

    2025

  34. [42]

    FLUX.2: Frontier visual intelligence,

    Black Forest Labs, “FLUX.2: Frontier visual intelligence,” https://bfl.ai/ blog/flux-2, 2025, blog announcement, November 25, 2025; introduces FLUX.2 [pro]

    2025

  35. [43]

    FLUX.1 Kontext: Flow matching for in-context image generation and editing in latent space,

    ——, “FLUX.1 Kontext: Flow matching for in-context image generation and editing in latent space,” 2025

    2025

  36. [44]

    Qwen-image technical report,

    Qwen Team, “Qwen-image technical report,” 2025, covers Qwen-Image and Qwen-Image-Edit

    2025

  37. [45]

    Gender shades: Intersectional accuracy disparities in commercial gender classification,

    J. Buolamwini and T. Gebru, “Gender shades: Intersectional accuracy disparities in commercial gender classification,” inConference on Fair- ness, Accountability and Transparency (FAccT), ser. PMLR, vol. 81, 2018, pp. 77–91

    2018

  38. [46]

    Model cards for model reporting,

    M. Mitchell, S. Wu, A. Zaldivar, P. Barnes, L. Vasserman, B. Hutchin- son, E. Spitzer, I. D. Raji, and T. Gebru, “Model cards for model reporting,” inConference on Fairness, Accountability, and Transparency (FAT*), 2019, pp. 220–229, arXiv:1810.03993

    2019 arXiv

Pith tools

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