Pith. sign in

REVIEW 3 major objections 5 minor 92 references

Video editing can be learned from synthetic video pairs generated by warping image-edit pairs, with no curated video data required.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 15:37 UTC pith:6A7H6Z5X

load-bearing objection A novel synthetic-data pipeline for video editing with an all-qualitative evaluation and a self-admitted caveat that undercuts the strongest reading of its central claim. the 3 major comments →

arxiv 2607.18227 v1 pith:6A7H6Z5X submitted 2026-07-20 cs.CV

FlowMimic: Mask-free Visual Editing and Generation with Pixel-pair Warped Flow Field for Online Video Editing Data Generation and Modality Mimicry

Dingyun Zhang , Lixue Gong , Wei Liu This is my paper
classification cs.CV
keywords video editingimage editingdata generationtemporal warped flow fieldmodality alignmentdiffusion transformerflow matchingreferring expression segmentation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper argues that video editing can be learned without any curated video editing data. The authors construct a pixel-pair temporal warped flow field: the same time-varying deformation, such as pan, zoom, rotation, stretch, or elastic motion, is applied to both source and target images of an existing image editing pair, producing a synthetic video pair in which every frame preserves the same pixel-level editing correspondence. They train a single image-video generation and editing model using only these online-generated video samples plus the original image samples, and report that it learns relighting, stylization, head swapping, object insertion, removal, virtual try-on, and other video editing tasks. They further claim that treating an image as a single-frame video and aligning the output distributions of the image and video modalities through modality-mimic generation and editing losses lets the two modalities improve each other. If this is right, the expensive mask annotation, image-to-video synthesis, and VLM-based filtering used in current video editing data pipelines become unnecessary, and any image editing task with loose layout correspondence can be scaled to video.

Core claim

The paper's central claim: temporal consistency of pixel-level editing correspondences, not naturalness of motion, is the ingredient needed to learn video editing. Applying one shared time-varying warp to source and target images of an image-edit pair yields a synthetic video pair in which every frame preserves the same pixel-pair edit relation; training on such pairs alone is reported to produce editing on real videos across relighting, stylization, swapping, insertion, removal, and try-on. Second, image is a single-frame video: a modality-mimic loss aligns text-to-image with text-to-video and image-editing with video-editing via KL divergence on one-step-denoised latents. Third, referring-

What carries the argument

The load-bearing object is the pixel-pair 4D temporal warped flow field: a canonical 3D grid of pixel coordinates for the source and target images is deformed by the same time-dependent warping, using a shared random seed, and each frame is produced by bilinear grid inverse sampling. Because the same deformation is applied to both images, every frame preserves the exact pixel-wise editing correspondence established in the first frame; the model sees a synchronized walk of edit pairs. Two supporting mechanisms carry the rest of the argument: modality-mimic losses, which compute KL divergence on softmaxed one-step reversed latents to align text-to-image with text-to-video and image-editing wit

Load-bearing premise

The load-bearing premise is that a model trained on synthetic videos made by applying the same parametric warp to image-edit pairs will transfer to real videos with natural object and camera motion, occlusions, lighting changes, and parallax; the paper asserts that temporal consistency of pixel-level editing correspondences suffices, but does not compare against real video editing data or natural-motion training baselines.

What would settle it

Train two models on the same image-edit pairs: one with only warped synthetic video pairs, and one with real video editing pairs spanning natural object motion, occlusions, and parallax. If the synthetic-only model performs comparably or better on held-out real videos, the sufficiency claim stands; if the real-data model is substantially better, the claim that temporal consistency of pixel-level editing correspondences suffices fails. A cheaper probe is to evaluate the synthetic-only model on videos whose motion types are absent from the warp family, for instance fast articulated motion with o

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Any image editing task with loose layout correspondence between source and target can be upgraded to video editing by applying the warp online during training, making video editing data a by-product of image editing data.
  • Training can start from a pretrained text-to-video model with a low learning rate and only a few thousand steps, so the data-generation overhead is minimal.
  • The model generalizes to longer videos than it saw in training, indicating the learned pixel-correspondence regularity is not tied to a fixed frame count.
  • Editing-region localization is learned from language and attention, so inference needs no mask sequence and no attached multimodal LLM.
  • Aligning image and video output distributions improves single-image generation realism and stylized video generation relative to the pretrained model.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the sufficiency claim holds, the scalability bottleneck for video editing shifts entirely to image editing data quality; improving datasets such as referring-expression masks or object-removal inpaints should directly improve video editing, a prediction that can be tested by dataset ablations.
  • The warp family acts as an implicit prior over motion; one could probe how far the model generalizes by evaluating on videos with articulated motion, occlusion, or parallax far outside the pan/zoom/rotation/stretch/elastic family, and by ablating motion types.
  • The modality-mimic losses exemplify a form of on-the-fly self-distillation in which the model's faster-converging image-editing pathway supervises its harder video-editing pathway and vice versa; this recipe may transfer to other paired modalities or favorable conditioning signals.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes FlowMimic, a post-training recipe built on the pretrained Wan2.1-T2V-1.3B model. It has three main components: (i) an online data-generation paradigm that converts image-editing pairs into video-editing pairs by applying a shared, time-varying parametric warp (pan, zoom, rotation, stretch, elastic, or compositions) to both source and target images; (ii) first-frame modality-mimic losses that align T2I with T2V and I2I with V2V via KL divergence between softmax-transformed, one-step-reversed latent estimates; and (iii) 'sense' tasks, chiefly referring-expression segmentation, with region-aware flow-matching and cross-attention losses. The central claim is that a model can learn video editing using only such online-generated warped video pairs, without specially curated real video-editing training data. The paper reports qualitative editing results on FiVE-Bench, UNIC-Bench, and additional web videos, along with cross-attention visualizations and stylized T2V examples.

Significance. If the central claim holds, the paper identifies a simple, scalable route to video-editing training data: temporal consistency of pixel-level edit correspondences, rather than natural motion or real-world coherence, may be sufficient. This would be a genuinely useful contribution, and the paper is unusually explicit about the limitations of its training data. The writing is also clear about the construction of the warp field. However, the evidence provided is almost entirely qualitative: no numeric metrics, no controlled ablation that isolates the flow field from the strong priors of the pretrained T2IV model, and no comparison against a real-video-pair baseline. The paper's own Fig. 13 caption attributes natural motion to the pretrained T2V/I2V capabilities, not to the proposed flow field. Consequently, the significance is currently prospective rather than established, and the strong 'only such data' claim is not yet supported.

major comments (3)
  1. [Sec. 3.3, Eqs. (38), (39), (42), (43)] The central claim—that a model can learn video editing using only online-generated warped image-pair data—is supported only by qualitative stills. There are no metrics on FiVE-Bench or UNIC-Bench, no comparison against a no-warp baseline, and no ablation that removes or varies the proposed flow field while keeping all other components fixed. Because Wan2.1-T2V already contains strong T2V and I2V motion priors, the qualitative results cannot establish that the pixel-pair warped flow field is the operative source of temporal editing competence. The Fig. 13 caption itself states that natural motion 'may arise from FlowMimic implicitly transferring the motion-generation ability—acquired from training on T2V and I2V tasks—to editing tasks', while the flow field is credited only with 'placement and consistency'. This directly weakens the abstract and Sec. 1 claim that video editing is learned
  2. [Sec. 3.4, Sec. 4.4, Fig. 28] The text claims that the modality-mimic editing loss enables 'mutual imitation' or 'bidirectional distribution alignment' because both distributions are produced by the same model. However, as implemented via torch.nn.functional.kl_div in Sec. 4.1, each KL term has a fixed target and gradients flow only through the 'input' distribution—for Eq. (39), the T2I path; for Eq. (43), the I2I path if log(p_i2i) is the input. The target distribution is not directly updated by the loss. Shared weights provide only an indirect coupling. Thus the loss as written is a one-directional regularizer, and the 'mutual mimic' claim is not a direct consequence of the loss geometry. Please specify the exact gradient flow, or use a symmetric divergence (e.g., JS divergence or Hellinger distance, both of which are mentioned as exploratory) if bidirectional alignment is intended.
  3. [Sec. 4.1, Sec. 5] The sense-related losses are a stated contribution, but their evaluation is limited to a few attention-map visualizations. No quantitative metrics are provided for referring-expression localization (e.g., mIoU or acc@0.5 on RefCOCO/RefCOCOg/gRefCOCO), no segmentation evaluation for the video referring-expression results, and no comparison of editing-region localization against the pretrained base model or against a model trained without the sense losses. Since 'the model has internalized this capability' is a load-bearing claim for the editing-region-aware component, the lack of quantitative validation is a major gap.
minor comments (5)
  1. [Sec. 3.2, Eqs. (5)–(29)] The implementation details give loss weights and warm-up schedules, but task sample ratios and per-task data weights are not specified beyond the 4:1 generation:editing ratio. Several tasks are described as receiving 'a prescribed ratio' or 'a modest weight'; these should be reported for reproducibility.
  2. [Sec. 3.4, Eq. (53)] The softmax is applied to the flattened latent of dimension D = (H/8)·(W/8)·16. This is an unusual choice; no justification or sensitivity analysis is given. If kept, please explain why the resulting distribution is semantically meaningful for KL alignment.
  3. [Throughout] The sense cross-attention loss uses the difference between mean attention inside and outside the ground-truth mask. This is reasonable, but the normalization by h_l·w_l in both terms cancels; consider reporting the actual normalization used.
  4. [Sec. 5] There are typographical and stylistic issues: 'embbedings', 'likewisely', 'JRK' rendering of template delimiters, and the very long timeline/attribution paragraph in Sec. 5. These do not affect the technical content but should be cleaned up.
  5. [Sec. 5] The paper explicitly acknowledges low-quality internal data, color discrepancies, and lack of per-task hyperparameter tuning. These are useful caveats, but they also underscore that the qualitative results may reflect favorable data selection; adding failure cases or quantitative error analysis would strengthen the report.

Circularity Check

0 steps flagged

No significant circularity: the flow-field data augmentation and mimic losses are training constructs, and the evidence for transfer is on out-of-distribution benchmarks rather than the fitted quantities.

full rationale

The central derivation—applying identical time-varying warps to source and target images (Sec. 3.2, Eqs. 5-28)—is a definitional data-generation scheme; it does not set a parameter from the evaluation set and then predict it. The modality mimic losses (Eqs. 38-43) are training objectives that directly minimize KL divergence; the paper's Sec. 4.4 observations of distributional similarity are on unseen FiVE-Bench/UNIC-Bench videos, so they are a generalization check, not the same fitted quantity. The sense losses are supervised by RefCOCO masks, and the attention-map visualizations (Fig. 28) are on UNIC-Bench, again outside the training distribution. The only self-citation (Rig-Your-Portrait, ref. [46]) is an anecdotal related-work note and is not load-bearing. The main weaknesses are evidential, not circular: evaluation is purely qualitative, and Fig. 13's caption explicitly credits natural motion to pretrained T2V/I2V priors rather than the flow field; these are correctness/evidence concerns, not reduction of outputs to inputs.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 0 invented entities

The central claim depends on hand-set motion strengths, loss weights, and task ratios, and on the unvalidated premise that synthetic pixel-correspondence training transfers to real video motion. The main training corpora are internal, so the full parameter and assumption count cannot be audited from the text alone.

free parameters (5)
  • motion strength hyperparameter mapping f_alpha(M) (pan distance, rotation angle, stretch intervals, elastic sigma/alpha0 = not specified; set to a moderate level
    Hand-chosen per motion type and said to control convergence speed and rigid vs non-rigid editing (Sec. 3.2 'Influence of Motion Strength'). No sensitivity analysis is reported.
  • loss weights (alpha_mimic_gen, alpha_mimic_editing, alpha_FM_SC, alpha_sense_attn) = 0.1, 0.1, 0.5, 1.0
    Given in Sec. 4.1 without ablation or tuning procedure; they determine the balance of all novel losses.
  • learning rate warmup schedule = 5e-6 to 1e-5 over 5000 steps
    Training hyperparameter (Sec. 4.1) not derived from first principles.
  • frame count distribution for online-generated video samples = uniform over {13, 25, 37, 49, 61}
    Chosen to fit memory constraints; the paper later tests 121-frame inference, but the effect of training length is not studied (Sec. 4.1).
  • task sample weights and ratios (generation:editing = 4:1, per-task weights) = not fully specified; 'prescribed ratio'
    Mixed-task behavior and convergence depend on these weights; the paper states they can be adjusted but does not give the full prescription (Sec. 4.3).
axioms (5)
  • standard math Flow matching linear interpolation and one-step reverse formula (Eqs. 1–2, 33–34) are valid for Wan2.1.
    Background from flow matching and Wan2.1; assumed valid rather than derived in this paper.
  • domain assumption Pixel-correspondence maintenance is sufficient for video-editing learning.
    Sec. 3.2: 'we posit that models do not require video pairs that conform to real-world coherence; rather, temporal consistency of pixel-level editing correspondences suffices.' This is the load-bearing generalization premise.
  • domain assumption Identical warps preserve the semantics of the edit.
    Sec. 3.2 claims 'the pixel-wise editing correspondence is preserved exactly'; for tasks like object insertion, warping an inserted object may distort its intended identity or shape. The paper assumes a 'loose layout correspondence' suffices.
  • domain assumption Reference elements as noise-free in-context conditioning (timestep 0, reference-inject self-attention mask) is a valid editing formulation.
    Sec. 3.3 and 3.5; this design choice is not independently validated, though it resembles prior in-context editing approaches.
  • ad hoc to paper Softmax over flattened one-step-reversed latents is a meaningful perceptual distribution for KL alignment.
    Sec. 3.3, Eqs. 36–39. The authors assert 'distributional shapes ... reflect the visual perceptual characteristics'; this is not established and is specific to the paper's loss design.

pith-pipeline@v1.3.0-alltime-deepseek · 57853 in / 13580 out tokens · 113021 ms · 2026-08-01T15:37:14.866099+00:00 · methodology

0 comments
read the original abstract

In line with the prevailing direction of vision research, we explore the integration of both generation and editing capabilities for video and image modalities within a single model. Current approaches to collecting video editing data typically depend on labour-intensive, time-consuming curated procedures--involving object mask annotation, the use of error-introducing pair synthesis via I2V model and ControlNet-like guidance, and VLM-based quality filtering or refinement--and demonstrate limited task scalability. As a result, the diversity of editing tasks remains substantially narrower than that available for image editing models. We develop a pixel-pair temporal warped flow field that can directly generate corresponding video editing samples in real time from image editing samples, and we demonstrate across multiple levels of video editing tasks that a model can learn video editing using only such data. We regard the image modality as a particular form of the video modality. Accordingly, we design a modality mimic generation loss and a modality mimic editing loss to relatively align the capabilities--and thereby the output distributions--of the two modalities through mutual imitation. Moreover, language-based visual editing entails the comprehension of the editing instruction and the reference visual content, the localization of the region corresponding to that instruction within the reference visual contents, and the modification of that region alone. Existing approaches predominantly rely on external aids, such as fine-tuning an additional MLLM or explicitly supplying a mask sequence as auxiliary input during inference. In contrast, we aspire for the model to internalize this capability. To that end, we introduce sense-related tasks--for instance, referring expression segmentation--along with corresponding editing-region-aware latent-level loss and attention-level loss.

Figures

Figures reproduced from arXiv: 2607.18227 by Dingyun Zhang, Lixue Gong, Wei Liu.

Figure 1
Figure 1. Figure 1: Overview of the video editing pair generation paradigm with pixel-pair temporal warped flow field. For any given image editing training sample of a task, we can obtain video editing samples in real time via this field that enable the model to learn the video editing capability for this task. Such video editing samples align with the model’s cognitive patterns, eliminating the need to adapt them to conform … view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the pretrained T2IV model [18] and our modality mimic generation loss. The area to the left of the grey dashed line provides a schematic overview of our adopted pretrained T2IV model, while the area to the right introduces the modality mimic generation loss used in our generation training step. The DiT blocks are shared components across both sides of the dashed line. In the schematic illustrat… view at source ↗
Figure 3
Figure 3. Figure 3: T2V first frame results from the pretrained T2IV model under increasing inference frame counts, using identical prompts and random seeds, at a resolution of 832 × 480. When the inference frame count is 1, the output corresponds to the T2I result. It can be observed that the T2I result exhibits a stronger synthetic or artificial appearance, with weaker cinematic realism compared to the first frame result fr… view at source ↗
Figure 4
Figure 4. Figure 4: Overview of the modality mimic editing loss in editing training step and our model architecture adaptations to editing task, including reference-inject self-attention mask, separate 3D rope with frame correspondence, and element identity encoding. In the schematic illustrating the modality mimic editing loss, for better clarity, we present the newly created I2I sample as the (M + 1)-th sample in the curren… view at source ↗
Figure 5
Figure 5. Figure 5: Convergence speed comparision between I2I and V2V. Given a reference video, I2I inference is applied to its first frame, while V2V inference is conducted on the complete video. A comparison is then performed between the first frames of the I2I and V2V inference outputs. The leftmost column presents the first frames of videos from FiVE-Bench [52]. The two central columns respectively exhibit the first frame… view at source ↗
Figure 6
Figure 6. Figure 6: Overview of sense-related tasks and losses. Additionally, the reference position select task is introduced in the bottom-right corner. Please refer to Sec. 3.4 and Sec. 3.5 for details, respectively. sequence to the DiT blocks. It should be noted that we regard the reference elements as conditional inputs, which, like the caption text embeddings, should provide guidance through￾out the denoising process an… view at source ↗
Figure 7
Figure 7. Figure 7 [PITH_FULL_IMAGE:figures/full_fig_p018_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Single-reference non-rigid element replacement results on FiVE-Bench [52]. Our method yields natural replacements while generating corresponding motion and physical effects, e.g., the walking motion of the lion in the second example and the dinosaur-tail shadow in the fourth. The sixth example demonstrates multi-target editing. averaging over this specific text embedding span: \bar {A}_{\text {ref}_1\text … view at source ↗
Figure 9
Figure 9. Figure 9: Single-reference rigid element replacement results on FiVE-Bench [52]. Our editing modifies only the reference region corresponding to the editing instruction, and the edited result exhibits a rigid correspondence with the reference object, as illustrated by the basketball in the third example. information from different semantic levels. The resulting 2D blocks-average cross-attention score map A¯ ref1-f q… view at source ↗
Figure 10
Figure 10. Figure 10: Single-reference color alteration and object addition results on FiVE-Bench [52]. Our editing results consistently incorporate corresponding physical effects, such as shadows. The second example demonstrates that our model can identify the editing region with occlusions (e.g., a backpack) and modify it exclusively. where ⊕ denotes the concatenation operation along the se￾quence length dimension. Separate … view at source ↗
Figure 11
Figure 11. Figure 11: Single-reference material modification results on FiVE-Bench [52]. Our method maintains pixel-aligned editing results across frames. index tx, height hx, and width wx. These angle sets are then applied to their respective channel groups via complex plane rotations. Formally, for a visual token x, the encoded output x˙ is obtained by: \begin {aligned} \dot {x} &= \operatorname {RoPE}(x, t_x, h_x, w_x) \\ &… view at source ↗
Figure 12
Figure 12. Figure 12: Object removal results on UNIC-Bench [22]. Unlike previous mainstream approaches [21, 64] that rely on object mask sequences to help the model locate regions for removal, we employ sense-related tasks and losses, enabling the model to delete specified objects through natural language. The fifth example demonstrates that our model can perceive and remove the associated physical effects, such as shadows, co… view at source ↗
Figure 13
Figure 13. Figure 13: Multi-reference object insertion results on UNIC-Bench [22]. Given an editing prompt and a reference image, FlowMimic can place the object from the reference image into the specified region of the reference video in a natural, prompt-following manner, while generating corresponding motions. It should be noted that we do not employ any specially constructed video editing data throughout the training proces… view at source ↗
Figure 14
Figure 14. Figure 14: Multi-reference object replacement results on UNIC-Bench [22]. Given an editing instruction that specifies the subject to be replaced and a reference image, FlowMimic can replace the subject in the reference video with the object from the reference image. When the subject in the reference video is in motion, as in the first and fourth examples, the edited video naturally preserves the corresponding motion… view at source ↗
Figure 15
Figure 15. Figure 15: Stylization and first frame propagation results on UNIC-Bench [22]. The stylization results demonstrate the model’s pixel-wise editing ability, whereas the first frame propagation results show that the editing applied to the initial frame can be propagated to later frames in a natural and physically plausible manner. Reference Position Select Task. In multi-reference edit￾ing tasks, the editing instructio… view at source ↗
Figure 16
Figure 16. Figure 16: Single-reference black-and-white video colorization, video deblurring, and video expanding results. The reference videos for deblurring are drawn from the test set of the GOPRO Large [65] dataset. of the selected target image within the set. The final sample is therefore represented as the tuple (\{I_{\text {ref}_1}, \dots , I_{\text {ref}_n}\}, I_\text {tar}) with the corresponding captions, and is used … view at source ↗
Figure 17
Figure 17. Figure 17: Video relighting, changing color tone, and video background replacement results. FlowMimic can change the lighting of the original scene and generate plausible physical effects in response to an editing instruction. When altering the color tone, the model faithfully preserves the layout of the original frames. In background replacement, the model also produces natural physical effects based on the editing… view at source ↗
Figure 18
Figure 18. Figure 18: Changing expression, changing hairstyle, and video virtual try-on results. Example three demonstrates that, even with substantial subject motion, our editing effect remains temporally consistent. and the target video receives an output caption constructed by prepending a template instruction to the original output caption. The template instruction takes a form such as: “Re￾ferring to the editing result of… view at source ↗
Figure 19
Figure 19. Figure 19: Video facial beautification, changing shape and stylization video to live-action video results. Here, “changing shape” refers to the alteration of the object’s objective form, such as its external age state or physical contour. The results of stylization video to live-action video demonstrate both the model’s capability to comprehend the semantics of non-photorealistic videos and its ability to preserve p… view at source ↗
Figure 20
Figure 20. Figure 20: Watermark removal, multi-reference material alteration and adding visual effect results. In the third and fourth examples, where the editing captions specify the bread and the girl, respectively, as the editing objects, the knife and the cup held by the girl remain unaltered. This demonstrates the model’s cross-modality semantic understanding and its region-aware editing ability. Generating visual effects… view at source ↗
Figure 21
Figure 21. Figure 21: Video face swapping and SAM3-like video referring expression segmentation results. Our face swapping and head swapping image editing samples each comprise only a few thousand instances and are of relatively low quality. For example, expressions, lip movements, gaze directions, and head poses are often inconsistent before and after the swap; nonetheless, the model demonstrates a degree of generalization ca… view at source ↗
Figure 22
Figure 22. Figure 22: Multi-subject and multi-task editing results. Although each editing training sample contains only a single-object, single-task editing, when the editing instruction contains multiple objects and various editing tasks at inference time, our model is also capable of identifying each object to be edited and executing the respective editing effects accordingly. This suggests that our model has fundamentally a… view at source ↗
Figure 23
Figure 23. Figure 23: Trajectory-guided conditional input editing. Our image editing tasks include both the condition map to image and its inverse operation, among which are approximately 3K image editing pairs—randomly sampled from the RefCOCO dataset—that involve generating a specified object from an image annotated with color-smeared blocks. We recognize that, for video editing tasks, this corresponding capability can give … view at source ↗
Figure 24
Figure 24. Figure 24: Generalizing to unseen conditional input task results. Our training data contain only a small number of samples related to normal maps, all of which are global normal maps of scenes with no data pertaining to bodies or portraits. Nevertheless, for the normal map to image task that involves only the foreground, the model displays some generalization capability. As shown in the second row, provided with a s… view at source ↗
Figure 25
Figure 25. Figure 25: Temporal generalization to more video frames during inference. During training, to conserve both CPU and GPU memory due to the online generation of video editing data, we generate videos with an average of 37 frames and a maximum of 61 frames. In inference, we investigate whether the model exhibits temporal generalization to longer sequences. Testing on 121 frames—within an acceptable inference time on th… view at source ↗
Figure 26
Figure 26. Figure 26: Comparison of the visual output distributions between video editing and image editing tasks under the same reference first frame. The modality mimic editing loss aims to align the editing capabilities between V2V and I2I. Given a reference video and an editing instruction, we perform video editing while also applying image editing to the first frame of the reference video. For the different tasks, the fir… view at source ↗
Figure 27
Figure 27. Figure 27: Comparison of T2I results between the pretrained T2IV model and our model under identical cinematic prompts. The modality mimic generation loss aims to align the generation distributions between T2V and T2I, bringing the latter closer to the former in terms of cinematic realism. The results indicate that, compared to the pretrained T2IV model, our model yields improved cinematic realism in T2I generation.… view at source ↗
Figure 28
Figure 28. Figure 28: Visualization comparisons of cross attention maps between referring expressions’ corresponding text embeddings and the first frame of the reference video on UNIC-Bench. To enhance the model’s comprehension of natural language and visual content, as well as its cross-modality matching capability, we design sense￾related tasks and losses. The results demonstrate that, compared with the pretrained T2IV model… view at source ↗
Figure 29
Figure 29. Figure 29: Comparison of T2V results on stylized prompts. For three common oil painting styles, we employ identical stylization prompts and random seeds for T2V generation. As can be seen from the results, compared to the pretrained T2IV model, our model exhibits a better stylistic response. tasks are readily available in large quantities from existing public datasets. 4.2. Benchmarks We present qualitative results … view at source ↗
Figure 30
Figure 30. Figure 30: In the first frame propagation test samples of UNIC-Bench [ [PITH_FULL_IMAGE:figures/full_fig_p049_30.png] view at source ↗
Figure 32
Figure 32. Figure 32: First frame propagation results with the color block abstraction method￾ology. It can be observed that the model is capable of generating plausible motion for the replaced object, despite distinct spatial discrepancies and action misalignment between it in the propagation frame and the original object in the first source frame. This holds true even when background inconsistencies are present due to editin… view at source ↗
Figure 33
Figure 33. Figure 33: Reference-to-Video (R2V) results with 61 frames and a resolution of 832 × 480 generated at inference. Given an editing prompt and reference images, our model demonstrates a degree of ability to generate the corresponding video in a natural, prompt-following manner, conditioned on the editing prompt and the visual information provided by the reference images. It should be noted that the training data of th… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

92 extracted references · 20 linked inside Pith

  1. [1]

    Consistent video- to-video transfer using synthetic dataset

    Jiaxin Cheng, Tianjun Xiao, and Tong He. Consistent video- to-video transfer using synthetic dataset. InICLR, 2026. 2, 4

    2026

  2. [2]

    Prompt-to-prompt image editing with cross attention control

    Amir Hertz, Ron Mokady, Jay Tenenbaum, Kfir Aberman, Yael Pritch, and Daniel Cohen-Or. Prompt-to-prompt image editing with cross attention control. InICLR, 2023. 4

    2023

  3. [3]

    Se ˜norita-2m: A high-quality instruction- based dataset for general video editing by video specialists

    Bojia Zi, Penghui Ruan, Marco Chen, Xianbiao Qi, Shaozhe Hao, Shihao Zhao, Youze Huang, Bin Liang, Rong Xiao, and Kam-Fai Wong. Se ˜norita-2m: A high-quality instruction- based dataset for general video editing by video specialists. InNeurIPS, 2025. 4

    2025

  4. [4]

    Scaling instruction-based video editing with a high-quality synthetic dataset

    Qingyan Bai, Qiuyu Wang, Hao Ouyang, Yue Yu, Hanlin Wang, Wen Wang, Ka Leong Cheng, Shuailei Ma, Yan- hong Zeng, Zichen Liu, et al. Scaling instruction-based video editing with a high-quality synthetic dataset. InarXiv preprint:2510.15742, 2025. 4

    arXiv 2025

  5. [5]

    Insvie-1m: Effective instruction-based video editing with elaborate dataset construction

    Yuhui Wu, Liyi Chen, Ruibin Li, Shihao Wang, Chenxi Xie, and Lei Zhang. Insvie-1m: Effective instruction-based video editing with elaborate dataset construction. InCVPR, 2025. 4

    2025

  6. [6]

    Openve-3m: A large-scale high-quality dataset for instruction- guided video editing

    Haoyang He, Jie Wang, Jiangning Zhang, Zhucun Xue, Xingyuan Bu, Qiangpeng Yang, Shilei Wen, and Lei Xie. Openve-3m: A large-scale high-quality dataset for instruction- guided video editing. InarXiv preprint:2512.07826, 2025. 2, 4

    arXiv 2025

  7. [7]

    A survey on multimodal large language models

    Shukang Yin, Chaoyou Fu, Sirui Zhao, Ke Li, Xing Sun, Tong Xu, and Enhong Chen. A survey on multimodal large language models. InNational Science Review, 2024. 2 5https://ipr.tsa.cn/

    2024

  8. [8]

    Vision-language models for vision tasks: A survey

    Jingyi Zhang, Jiaxing Huang, Sheng Jin, and Shijian Lu. Vision-language models for vision tasks: A survey. InTPAMI,

  9. [9]

    A survey of state of the art large vision language models: Benchmark evaluations and challenges

    Zongxia Li, Xiyang Wu, Hongyang Du, Fuxiao Liu, Huy Nghiem, and Guangyao Shi. A survey of state of the art large vision language models: Benchmark evaluations and challenges. InCVPR, 2025. 2

    2025

  10. [10]

    Sam 2: Segment anything in images and videos

    Nikhila Ravi, Valentin Gabeur, Yuan-Ting Hu, Ronghang Hu, Chaitanya Ryali, Tengyu Ma, Haitham Khedr, Roman R¨adle, Chloe Rolland, Laura Gustafson, et al. Sam 2: Segment anything in images and videos. InICLR, 2025. 2, 15

    2025

  11. [11]

    Sam 3: Segment anything with concepts

    Nicolas Carion, Laura Gustafson, Yuan-Ting Hu, Shoub- hik Debnath, Ronghang Hu, Didac Suris, Chaitanya Ryali, Kalyan Vasudev Alwala, Haitham Khedr, Andrew Huang, et al. Sam 3: Segment anything with concepts. InarXiv preprint:2511.16719, 2025. 2, 15

    Pith/arXiv arXiv 2025

  12. [12]

    Grounding dino: Marrying dino with grounded pre-training for open-set object detection

    Shilong Liu, Zhaoyang Zeng, Tianhe Ren, Feng Li, Hao Zhang, Jie Yang, Qing Jiang, Chunyuan Li, Jianwei Yang, Hang Su, et al. Grounding dino: Marrying dino with grounded pre-training for open-set object detection. InECCV, 2024. 2

    2024

  13. [13]

    Qwen2-vl: Enhancing vision-language model’s perception of the world at any resolution

    Peng Wang, Shuai Bai, Sinan Tan, Shijie Wang, Zhihao Fan, Jinze Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, et al. Qwen2-vl: Enhancing vision-language model’s perception of the world at any resolution. InarXiv preprint:2409.12191, 2024. 2, 15

    Pith/arXiv arXiv 2024

  14. [14]

    Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wen- bin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, et al. Qwen2. 5-vl technical report. InarXiv preprint:2502.13923, 2025

    Pith/arXiv arXiv 2025

  15. [15]

    Qwen3-vl technical report

    Shuai Bai, Yuxuan Cai, Ruizhe Chen, Keqin Chen, Xionghui Chen, Zesen Cheng, Lianghao Deng, Wei Ding, Chang Gao, Chunjiang Ge, et al. Qwen3-vl technical report. InarXiv preprint:2511.21631, 2025. 2, 15

    Pith/arXiv arXiv 2025

  16. [16]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. In ICCV, 2023. 2, 5

    2023

  17. [17]

    Reward models in deep reinforcement learning: A survey

    Rui Yu, Shenghua Wan, Yucen Wang, Chen-Xiao Gao, Le Gan, Zongzhang Zhang, and De-Chuan Zhan. Reward models in deep reinforcement learning: A survey. InIJCAI,

  18. [18]

    Wan: Open and advanced large-scale video gener- ative models

    Team Wan, Ang Wang, Baole Ai, Bin Wen, Chaojie Mao, Chen-Wei Xie, Di Chen, Feiwu Yu, Haiming Zhao, Jianxiao Yang, et al. Wan: Open and advanced large-scale video gener- ative models. InarXiv preprint:2503.20314, 2025. 2, 3, 4, 5, 6, 18, 38

    Pith/arXiv arXiv 2025

  19. [19]

    Instructx: To- wards unified visual editing with mllm guidance

    Chong Mou, Qichao Sun, Yanze Wu, Pengze Zhang, Xinghui Li, Fulong Ye, Songtao Zhao, and Qian He. Instructx: To- wards unified visual editing with mllm guidance. InarXiv preprint:2510.08485, 2025. 2, 3, 6, 13

    arXiv 2025

  20. [20]

    Vino: A unified visual generator with interleaved omnimodal context

    Junyi Chen, Tong He, Zhoujie Fu, Pengfei Wan, Kun Gai, and Weicai Ye. Vino: A unified visual generator with interleaved omnimodal context. InarXiv preprint:2601.02358, 2026. 2, 3, 6, 13

    arXiv 2026

  21. [21]

    Videopainter: Any- length video inpainting and editing with plug-and-play con- text control

    Yuxuan Bian, Zhaoyang Zhang, Xuan Ju, Mingdeng Cao, Liangbin Xie, Ying Shan, and Qiang Xu. Videopainter: Any- length video inpainting and editing with plug-and-play con- text control. InSIGGRAPH, 2025. 3, 4, 24

    2025

  22. [22]

    Unic: Unified in-context video editing

    Zixuan Ye, Xuanhua He, Quande Liu, Qiulin Wang, Xintao Wang, Pengfei Wan, Di Zhang, Kun Gai, Qifeng Chen, and Wenhan Luo. Unic: Unified in-context video editing. InarXiv preprint:2506.04216, 2025. 4, 13, 24, 25, 26, 27, 30, 41, 49

    Pith/arXiv arXiv 2025

  23. [23]

    Anyportal: Zero-shot consistent video background replacement

    Wenshuo Gao, Xicheng Lan, and Shuai Yang. Anyportal: Zero-shot consistent video background replacement. InICCV,

  24. [24]

    Generative video propagation

    Shaoteng Liu, Tianyu Wang, Jui-Hsien Wang, Qing Liu, Zhifei Zhang, Joon-Young Lee, Yijun Li, Bei Yu, Zhe Lin, Soo Ye Kim, et al. Generative video propagation. InCVPR,

  25. [25]

    Univideo: Unified understanding, generation, and editing for videos

    Cong Wei, Quande Liu, Zixuan Ye, Qiulin Wang, Xintao Wang, Pengfei Wan, Kun Gai, and Wenhu Chen. Univideo: Unified understanding, generation, and editing for videos. In ICLR, 2026. 3, 6, 13

    2026

  26. [26]

    Magicbrush: A manually annotated dataset for instruction- guided image editing

    Kai Zhang, Lingbo Mo, Wenhu Chen, Huan Sun, and Yu Su. Magicbrush: A manually annotated dataset for instruction- guided image editing. InNeurIPS, 2023. 4

    2023

  27. [27]

    Pico-banana- 400k: A large-scale dataset for text-guided image editing

    Yusu Qian, Eli Bocek-Rivele, Liangchen Song, Jialing Tong, Yinfei Yang, Jiasen Lu, Wenze Hu, and Zhe Gan. Pico-banana- 400k: A large-scale dataset for text-guided image editing. arXiv preprint:2510.19808, 2025. 4

    arXiv 2025

  28. [28]

    Gpt-image-edit-1.5 m: A million-scale, gpt-generated image dataset

    Yuhan Wang, Siwei Yang, Bingchen Zhao, Letian Zhang, Qing Liu, Yuyin Zhou, and Cihang Xie. Gpt-image-edit-1.5 m: A million-scale, gpt-generated image dataset. InarXiv preprint:2507.21033, 2025. 4

    Pith/arXiv arXiv 2025

  29. [29]

    Imgedit: A unified image editing dataset and benchmark

    Yang Ye, Xianyi He, Zongjian Li, Bin Lin, Shenghai Yuan, Zhiyuan Yan, Bohan Hou, and Li Yuan. Imgedit: A unified image editing dataset and benchmark. InarXiv preprint:2505.20275, 2025. 4

    Pith/arXiv arXiv 2025

  30. [30]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. InICML, 2021. 4, 54

    2021

  31. [31]

    Relightmaster: Precise video relighting with multi-plane light images

    Weikang Bian, Xiaoyu Shi, Zhaoyang Huang, Jianhong Bai, Qinghe Wang, Xintao Wang, Pengfei Wan, Kun Gai, and Hongsheng Li. Relightmaster: Precise video relighting with multi-plane light images. InarXiv preprint:2511.06271, 2025. 4

    arXiv 2025

  32. [32]

    Recammaster: Camera-controlled generative ren- dering from a single video

    Jianhong Bai, Menghan Xia, Xiao Fu, Xintao Wang, Lianrui Mu, Jinwen Cao, Zuozhu Liu, Haoji Hu, Xiang Bai, Pengfei Wan, et al. Recammaster: Camera-controlled generative ren- dering from a single video. InICLR, 2026. 4

    2026

  33. [33]

    Anyv2v: A tuning-free framework for any video-to- video editing tasks

    Max Ku, Cong Wei, Weiming Ren, Harry Yang, and Wenhu Chen. Anyv2v: A tuning-free framework for any video-to- video editing tasks. InTMLR, 2024. 4, 28

    2024

  34. [34]

    Denoising diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. InICLR, 2021. 4

    2021

  35. [35]

    Videoshop: Localized semantic video editing with noise-extrapolated dif- fusion inversion

    Xiang Fan, Anand Bhattad, and Ranjay Krishna. Videoshop: Localized semantic video editing with noise-extrapolated dif- fusion inversion. InECCV, 2024. 4

    2024

  36. [36]

    Con- textflow: Training-free video object editing via adaptive con- text enrichment

    Yiyang Chen, Xuanhua He, Xiujun Ma, and Jack Ma. Con- textflow: Training-free video object editing via adaptive con- text enrichment. InAAAI, 2026. 4, 19

    2026

  37. [37]

    Flow matching for generative modeling

    Yaron Lipman, Ricky TQ Chen, Heli Ben-Hamu, Maximilian Nickel, and Matt Le. Flow matching for generative modeling. InICLR, 2023. 4, 6

    2023

  38. [38]

    Lora: Low-rank adaptation of large language models

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Liang Wang, Weizhu Chen, et al. Lora: Low-rank adaptation of large language models. InICLR, 2022. 4

    2022

  39. [39]

    I2vedit: First-frame-guided video editing via image-to- video diffusion models

    Wenqi Ouyang, Yi Dong, Lei Yang, Jianlou Si, and Xingang Pan. I2vedit: First-frame-guided video editing via image-to- video diffusion models. InSIGGRAPH Asia, 2024. 4

    2024

  40. [40]

    Vace: All-in-one video creation and editing

    Zeyinzi Jiang, Zhen Han, Chaojie Mao, Jingfeng Zhang, Yulin Pan, and Yu Liu. Vace: All-in-one video creation and editing. InICCV, 2025. 4

    2025

  41. [41]

    Scalable diffusion models with transformers

    William Peebles and Saining Xie. Scalable diffusion models with transformers. InICCV, 2023. 4, 6

    2023

  42. [42]

    Roformer: Enhanced transformer with rotary position embedding

    Jianlin Su, Murtadha Ahmed, Yu Lu, Shengfeng Pan, Wen Bo, and Yunfeng Liu. Roformer: Enhanced transformer with rotary position embedding. InNeurocomputing, 2024. 4, 22

    2024

  43. [43]

    Light-a-video: Training-free video relight- ing via progressive light fusion

    Yujie Zhou, Jiazi Bu, Pengyang Ling, Pan Zhang, Tong Wu, Qidong Huang, Jinsong Li, Xiaoyi Dong, Yuhang Zang, Yuhang Cao, et al. Light-a-video: Training-free video relight- ing via progressive light fusion. InICCV, 2025. 5

    2025

  44. [44]

    Vface: A training-free approach for diffusion-based video face swapping

    Sanoojan Baliah, Yohan Abeysinghe, Rusiru Thushara, Khan Muhammad, Abhinav Dhall, Karthik Nandakumar, and Muhammad Haris Khan. Vface: A training-free approach for diffusion-based video face swapping. InWACV, 2026. 5

    2026

  45. [45]

    The devil is in the details: Enhancing video virtual try-on via keyframe-driven details injection

    Qingdong He, Xueqin Chen, Yanjie Pan, Peng Tang, Pengcheng Xu, Zhenye Gan, Chengjie Wang, Xiaobin Hu, Jiangning Zhang, and Yabiao Wang. The devil is in the details: Enhancing video virtual try-on via keyframe-driven details injection. InarXiv preprint:2512.20340, 2025. 5

    Pith/arXiv arXiv 2025

  46. [46]

    Rig-your-portrait: Controllable and re- lightable portrait video generation with explicit 3d guidance

    Dingyun Zhang. Rig-your-portrait: Controllable and re- lightable portrait video generation with explicit 3d guidance. https://rigyourportrait.github.io , 2025. 5, 42

    2025

  47. [47]

    Editverse: Unifying image and video editing and generation with in-context learning

    Xuan Ju, Tianyu Wang, Yuqian Zhou, He Zhang, Qing Liu, Nanxuan Zhao, Zhifei Zhang, Yijun Li, Yuanhao Cai, Shaoteng Liu, et al. Editverse: Unifying image and video editing and generation with in-context learning. InarXiv preprint:2509.20360, 2025. 6

    Pith/arXiv arXiv 2025

  48. [48]

    Omni-video 2: Scaling mllm-conditioned dif- fusion for unified video generation and editing

    Hao Yang, Zhiyu Tan, Jia Gong, Luozheng Qin, Hesen Chen, Xiaomeng Yang, Yuqing Sun, Yuetan Lin, Mengping Yang, and Hao Li. Omni-video 2: Scaling mllm-conditioned dif- fusion for unified video generation and editing. InarXiv preprint:2602.08820, 2026. 6

    arXiv 2026

  49. [49]

    Multimodal referring segmentation: A survey

    Henghui Ding, Song Tang, Shuting He, Chang Liu, Zuxuan Wu, and Yu-Gang Jiang. Multimodal referring segmentation: A survey. InarXiv preprint:2508.00265, 2025. 6

    Pith/arXiv arXiv 2025

  50. [50]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszko- reit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. InNIPS, 2017. 6

    2017

  51. [51]

    Exploring the limits of transfer learning with a unified text-to-text transformer

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. Exploring the limits of transfer learning with a unified text-to-text transformer. InJMLR, 2020. 6, 17

    2020

  52. [52]

    Five-bench: A fine-grained video editing benchmark for evaluating emerging diffusion and rectified flow models

    Minghan Li, Chenxi Xie, Yichen Wu, Lei Zhang, and Mengyu Wang. Five-bench: A fine-grained video editing benchmark for evaluating emerging diffusion and rectified flow models. InICCV, 2025. 13, 19, 20, 22, 23, 41

    2025

  53. [53]

    Sparse videogen: Accelerating video dif- fusion transformers with spatial-temporal sparsity

    Haocheng Xi, Shuo Yang, Yilong Zhao, Chenfeng Xu, Muyang Li, Xiuyu Li, Yujun Lin, Han Cai, Jintao Zhang, Dacheng Li, et al. Sparse videogen: Accelerating video dif- fusion transformers with spatial-temporal sparsity. InICML,

  54. [54]

    Kullback-leibler divergence

    James M Joyce. Kullback-leibler divergence. InInternational encyclopedia of statistical science, 2025. 13

    2025

  55. [55]

    Unireal: Universal image generation and editing via learning real-world dynamics

    Xi Chen, Zhifei Zhang, He Zhang, Yuqian Zhou, Soo Ye Kim, Qing Liu, Yijun Li, Jianming Zhang, Nanxuan Zhao, Yilin Wang, et al. Unireal: Universal image generation and editing via learning real-world dynamics. InCVPR, 2025. 13

    2025

  56. [56]

    Omnigen: Unified image generation

    Shitao Xiao, Yueze Wang, Junjie Zhou, Huaying Yuan, Xin- grun Xing, Ruiran Yan, Chaofan Li, Shuting Wang, Tiejun Huang, and Zheng Liu. Omnigen: Unified image generation. InCVPR, 2025. 13, 50

    2025

  57. [57]

    Language as queries for referring video object segmentation

    Jiannan Wu, Yi Jiang, Peize Sun, Zehuan Yuan, and Ping Luo. Language as queries for referring video object segmentation. InCVPR, 2022. 16

    2022

  58. [58]

    Semantic and sequential alignment for referring video object segmentation

    Feiyu Pan, Hao Fang, Fangkai Li, Yanyu Xu, Yawei Li, Luca Benini, and Xiankai Lu. Semantic and sequential alignment for referring video object segmentation. InCVPR, 2025. 16

    2025

  59. [59]

    Modeling context in referring expressions

    Licheng Yu, Patrick Poirson, Shan Yang, Alexander C Berg, and Tamara L Berg. Modeling context in referring expressions. InECCV, 2016. 16, 40

    2016

  60. [60]

    Referitgame: Referring to objects in pho- tographs of natural scenes

    Sahar Kazemzadeh, Vicente Ordonez, Mark Matten, and Tamara Berg. Referitgame: Referring to objects in pho- tographs of natural scenes. InEMNLP, 2014. 16

    2014

  61. [61]

    Gres: General- ized referring expression segmentation

    Chang Liu, Henghui Ding, and Xudong Jiang. Gres: General- ized referring expression segmentation. InCVPR, 2023. 16, 40

    2023

  62. [62]

    Fastcomposer: Tuning-free multi- subject image generation with localized attention

    Guangxuan Xiao, Tianwei Yin, William T Freeman, Fr ´edo Durand, and Song Han. Fastcomposer: Tuning-free multi- subject image generation with localized attention. InIJCV,

  63. [63]

    Do vision trans- formers see like convolutional neural networks? InNeurIPS,

    Maithra Raghu, Thomas Unterthiner, Simon Kornblith, Chiyuan Zhang, and Alexey Dosovitskiy. Do vision trans- formers see like convolutional neural networks? InNeurIPS,

  64. [64]

    From ideal to real: Stable video object removal under imperfect conditions

    Jiagao Hu, Yuxuan Chen, Fuhao Li, Zepeng Wang, Fei Wang, Daiguo Zhou, and Jian Luan. From ideal to real: Stable video object removal under imperfect conditions. InarXiv preprint:2603.09283, 2026. 24

    Pith/arXiv arXiv 2026

  65. [65]

    Deep multi-scale convolutional neural network for dynamic scene deblurring

    Seungjun Nah, Tae Hyun Kim, and Kyoung Mu Lee. Deep multi-scale convolutional neural network for dynamic scene deblurring. InCVPR, 2017. 28, 42

    2017

  66. [66]

    First frame is the place to go for video content customization

    Jingxi Chen, Zongxia Li, Zhichao Liu, Guangyao Shi, Xiyang Wu, Fuxiao Liu, Cornelia Fermuller, Brandon Y Feng, and Yiannis Aloimonos. First frame is the place to go for video content customization. InarXiv preprint:2511.15700, 2025. 24

    arXiv 2025

  67. [67]

    Deep learning-based image and video inpainting: A survey

    Weize Quan, Jiaxi Chen, Yanli Liu, Dong-Ming Yan, and Pe- ter Wonka. Deep learning-based image and video inpainting: A survey. InIJCV, 2024. 34

    2024

  68. [68]

    Unboxed: Geometrically and tem- porally consistent video outpainting

    Zhongrui Yu, Martina Megaro-Boldini, Robert W Sumner, and Abdelaziz Djelouah. Unboxed: Geometrically and tem- porally consistent video outpainting. InCVPR, 2025. 35

    2025

  69. [69]

    Colorflux: A structure- color decoupling framework for old photo colorization

    Bingchen Li, Zhixin Wang, Fan Li, Jiaqi Xu, Jiaming Guo, Renjing Pei, Xin Li, and Zhibo Chen. Colorflux: A structure- color decoupling framework for old photo colorization. In arXiv preprint:2603.28162, 2026. 37

    arXiv 2026

  70. [70]

    Colorsurge: Bringing vibrancy and efficiency to automatic video colorization via dual-branch fusion

    Hongbo Zhao, Jiaxing Li, Peiyi Zhang, Peng Xiao, Jianxin Lin, and Yijun Wang. Colorsurge: Bringing vibrancy and efficiency to automatic video colorization via dual-branch fusion. InSIGGRAPH, 2025. 37

    2025

  71. [71]

    Objectclear: Complete object removal via object-effect attention

    Jixin Zhao, Shangchen Zhou, Zhouxia Wang, Peiqing Yang, and Chen Change Loy. Objectclear: Complete object removal via object-effect attention. InCVPR, 2026. 40

    2026

  72. [72]

    Insert anything: Image insertion via in-context editing in dit

    Wensong Song, Hong Jiang, Zongxin Yang, Zheqiao Cheng, Ruijie Quan, and Yi Yang. Insert anything: Image insertion via in-context editing in dit. InAAAI, 2026. 40

    2026

  73. [73]

    Champ: Controllable and consistent human image animation with 3d parametric guidance

    Shenhao Zhu, Junming Leo Chen, Zuozhuo Dai, Zilong Dong, Yinghui Xu, Xun Cao, Yao Yao, Hao Zhu, and Siyu Zhu. Champ: Controllable and consistent human image animation with 3d parametric guidance. InECCV, 2024. 42

    2024

  74. [74]

    Phantom-data: Towards a general subject-consistent video generation dataset

    Zhuowei Chen, Bingchuan Li, Tianxiang Ma, Lijie Liu, Ming- cong Liu, Yi Zhang, Gen Li, Xinghui Li, Siyu Zhou, Qian He, et al. Phantom-data: Towards a general subject-consistent video generation dataset. InarXiv preprint:2506.18851, 2025. 50

    Pith/arXiv arXiv 2025

  75. [75]

    Phan- tom: Subject-consistent video generation via cross-modal alignment

    Lijie Liu, Tianxiang Ma, Bingchuan Li, Zhuowei Chen, Jiawei Liu, Gen Li, Siyu Zhou, Qian He, and Xinglong Wu. Phan- tom: Subject-consistent video generation via cross-modal alignment. InICCV, 2025. 50, 54

    2025

  76. [76]

    Self-distilled reasoner: On-policy self-distillation for large language models

    Siyan Zhao, Zhihui Xie, Mengchen Liu, Jing Huang, Guan Pang, Feiyu Chen, and Aditya Grover. Self-distilled reasoner: On-policy self-distillation for large language models. InarXiv preprint:2601.18734, 2026. 50

    Pith/arXiv arXiv 2026

  77. [77]

    A survey of on- policy distillation for large language models

    Mingyang Song and Mao Zheng. A survey of on- policy distillation for large language models. InarXiv preprint:2604.00626, 2026

    Pith/arXiv arXiv 2026

  78. [78]

    Reinforcement learning: A survey

    Leslie Pack Kaelbling, Michael L Littman, and Andrew W Moore. Reinforcement learning: A survey. InJAIR, 1996. 50

    1996

  79. [79]

    Z-image: An efficient image generation foundation model with single-stream diffusion transformer

    Huanqia Cai, Sihan Cao, Ruoyi Du, Peng Gao, Steven Hoi, Zhaohui Hou, Shijie Huang, Dengyang Jiang, Xin Jin, Liangchen Li, et al. Z-image: An efficient image generation foundation model with single-stream diffusion transformer. arXiv preprint:2511.22699, 2025. 50

    Pith/arXiv arXiv 2025

  80. [80]

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

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim En- tezari, Jonas M¨uller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, et al. Scaling rectified flow trans- formers for high-resolution image synthesis. InICML, 2024. 50

    2024

Showing first 80 references.