Human Action CLIPs: Detecting AI-generated Human Motion

Hany Farid; Matyas Bohacek

REVIEW 4 major objections 6 minor 1 cited by

Human Action CLIPs: Detecting AI-generated Human Motion

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Semantic embeddings that align video frames with text can tell real from AI-generated human motion with up to 99.2% video-level accuracy.

desk verdict Solid dataset and cross-generator results for CLIP-based video deepfake detection, but the abstract overstates resolution robustness and the 'unsupervised' frame-to-prompt claim needs an honest rewrite. read the letter →

arxiv 2412.00526 v2 pith:SU6PZTAJ submitted 2024-11-30 cs.CV cs.AIcs.GR

Matyas Bohacek , Hany Farid This is my paper · ORCID

classification cs.CVcs.AIcs.GR

keywords AI-generatedvideodetectiondeepfakeCLIPembeddingsmediaforensicstext-to-videohumanmotionsemanticDeepActiondataset

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that the semantic meaning of a video frame, as captured by multi-modal embeddings that align images with text, is enough to tell real footage of people from AI-generated footage of people. The authors build DeepAction, a dataset of 3,100 AI clips from seven text-to-video generators plus 100 matching real clips, and show that simple classifiers on these embeddings reach video-level accuracy as high as 99.2%. The signal survives heavy resolution and compression laundering, and it transfers to generators the classifier never saw, averaging 95.2% accuracy in leave-one-out tests. Even a training-free check that compares a frame to the text prompts 'real image' and 'fake image' reaches 96.2% video accuracy. If this holds, a lightweight semantic detector could be deployed quickly and keep working as generators improve.

What carries the argument

The load-bearing mechanism is the multi-modal semantic embedding: a neural network (CLIP or its variants SigLIP and JinaCLIP) that maps an image and a text caption into a shared vector space, so a video frame can be compared directly to words. The authors add a fine-tuned CLIP (FT-CLIP) trained on real/fake frame pairs captioned as 'a REAL image of {action}' versus 'a FAKE image of {action}', which sharpens the real/fake axis. Classification is deliberately simple: a support-vector machine over per-frame embeddings (two-class 'real vs fake' or eight-class 'which generator'), with a video labeled fake when most frames are fake, plus a training-free 'frame-to-prompt' classifier that assigns a frame to whichever of two text prompts ('authentic image' vs 'AI-generated image') has the higher cosine similarity. The DeepAction dataset, 3,100 AI clips from seven generators matched to 100 real stock clips on the action prompt, is the testbed that lets the authors measure both accuracy and generalization.

What would settle it

Train the two-class classifier on DeepAction, then test it on a fresh collection of real videos filmed with varied cameras, lighting, and compression levels, and on clips from a newly released text-to-video model not among the seven generators. If video-level accuracy for the real videos falls well below the reported 98.5% or for the unseen model below roughly 90%, the claimed robustness and generalization would be contradicted.

Watch

Extended reading notes

Core claim

The central discovery is that the multi-modal embedding space carries a dependable real-versus-fake axis for generated video, even though the embeddings were never trained for forensics. On the DeepAction dataset, a fine-tuned CLIP model followed by a polynomial-kernel SVM reaches 99.2% video-level accuracy, and a fully unsupervised cosine-similarity comparison to the text prompts 'real image' and 'fake image' reaches 96.2%. The separation survives dropping resolution to 25% and bitrate to 10%, and it transfers to a generator withheld from training with an average video accuracy of 95.2%. The same signal also flags 97.1% of non-human AI-generated videos as fake, while failing on talking-head deepfakes (55.8%) and CGI (39.1%). The authors read this as evidence that semantic embeddings capture a general, content-agnostic signature of whole-scene generative synthesis rather than a pixel-level artifact.

Load-bearing premise

The 100 real stock videos and 3,100 AI clips from seven generators are matched only on the action prompt, so the reported accuracy assumes these two collections represent how real and AI-generated human-motion videos actually look in the wild.

Editorial extensions

If this is right

Semantic-embedding classification keeps video-level accuracy above 90% even when the bitrate is cut to 10% of the original, so the detector can survive typical social-media re-encoding.
In leave-one-out tests, a classifier trained on six generators detects the seventh at an average video accuracy of 95.2%, so the detector is not tied to the quirks of one synthesis model.
The training-free frame-to-prompt variant reaches 96.2% video accuracy, so a usable detector could be deployed without collecting labeled real/fake pairs.
The same model flags 97.1% of AI-generated videos without humans as fake, suggesting the embedding captures a general sign of generative synthesis rather than a human-motion-specific artifact.
The detector does not transfer to talking-head deepfakes (55.8%) or CGI (39.1%), so it is a detector of full-body generated video, not a universal synthetic-video detector.

Reading between the lines

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

The authors do not say this, but the frame-to-prompt result implies the text prompts themselves define a real/fake axis in the embedding space; one testable extension is to use other prompt pairs (for example 'natural motion' versus 'synthetic motion') and see whether accuracy changes.
Because the method works on non-human content, the underlying signal may be a shared property of diffusion or latent-video generation rather than of human bodies; an extension would be to test the same embeddings on AI-generated audio spectrograms or synthetic 3D scenes.
The authors leave open whether adversarial perturbations in embedding space can fool the classifier; a concrete extension is to run white-box attacks on the SVM decision boundary to measure how much perturbation flips a fake video to 'real'.
Since the real clips come from a single stock-video source, deploying the detector on user-generated video with different cameras, motion blur, and stabilization would test whether the 98.5% real-video accuracy holds beyond clean Pexels footage.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Quick take: this is a legitimate empirical contribution—a new 3,100-clip dataset of text-to-video human motion from seven generators plus matched real footage, a clean comparison of four CLIP-family embeddings, and the best evidence yet that semantic embeddings generalize across unseen generators. The core supervised detection is solid. But the abstract oversells resolution robustness, and the 'unsupervised' frame-to-prompt method is not as clean as advertised. Neither problem sinks the paper; both need fixing before publication.

What's new: DeepAction itself is the main gift. Evaluating on action-level held-out splits with macro-averaged accuracy, five repetitions, and leave-one-out across generators is a step beyond the image-only CLIP deepfake papers. The FT-CLIP variant reaching 99.2% video accuracy on the in-distribution test is credible, and the leave-one-out 91.9–96.7% range is the strongest result in the paper. The compression robustness test is also genuinely good—92.3% at 10% bitrate.

Soft spots: - The resolution-robustness claim in the abstract does not match what is tested. Table 3 only uses the generic CLIP embedding with a linear SVM; at 25% resolution video accuracy drops to 72.8%, and at 10% it is at chance. The authors' best FT-CLIP model is never evaluated under resolution reduction. The stress-test note is right: as written, 'robust to ... resolution attacks' is unsupported and partly contradicted by the paper's own table. This is fixable by testing the best method and by softening the claim. - The frame-to-prompt method is described as requiring no explicit training, but the only version that works (95.2% frame accuracy) uses FT-CLIP, fine-tuned on the same DeepAction dataset. With off-the-shelf CLIP, frame-to-prompt is at chance. Call it 'prompt-based classification with a fine-tuned encoder' or report the fine-tuned variant honestly as supervised in the embedding. - No code is released. The dataset is public, which is good; for a method paper, releasing the SVM training pipeline or at least embeddings would materially help replication. - Evaluation is careful but small on the real side: 100 Pexels videos. No error bars or confidence intervals on the five repetitions. I'd want those before relying on the 99.2% number.

The central argument—that semantic embeddings separate real from AI-generated human motion and generalize to new generators reasonably well—holds up. The robustness and no-training claims are overstated, not fabricated.

This paper deserves a serious referee. Send it out, but expect a major revision on framing. I'd bring it to reading group and would cite the dataset.

Referee Report

4 major / 6 minor

Summary. The paper introduces DeepAction, a new open-sourced dataset of 3,100 AI-generated video clips produced by seven text-to-video models together with 100 real Pexels videos matched on 100 human-action prompts. The authors evaluate four multi-modal embedding models (CLIP, SigLIP, JinaCLIP, and a custom fine-tuned CLIP) with linear, RBF, and polynomial SVM classifiers for two-class real/fake detection and eight-class source attribution, as well as a 'frame-to-prompt' cosine-similarity classifier that requires no classifier training. The paper reports video-level two-class accuracies up to 99.2%, an average leave-one-out generalization accuracy of 95.2% for the CLIP linear SVM, robustness to compression, and mixed results on resolution, non-human motion, talking-head deepfakes, and CGI.

Significance. If the reported results hold, the paper demonstrates that simple linear classifiers on multi-modal semantic embeddings provide a strong baseline for detecting AI-generated human-motion video, and the release of DeepAction is a useful benchmark for the community. Strengths include the action-level held-out split, macro-averaged evaluation over five repetitions, leave-one-out model generalization analysis, and evaluation on external datasets (DeepSpeak, GTA-Human, and Sora/Runway non-human videos). However, several load-bearing claims are currently overstated, particularly the abstract's robustness claim, the 'unsupervised' description of frame-to-prompt classification, and the non-human generalization result, which lacks a real-class control.

major comments (4)

[Abstract and §5.1 (Table 3)] The abstract's claim that the method is 'robust' to laundering 'including resolution and compression attacks' is not supported by the paper's own evaluation. Table 3 shows that the only resolution-robustness experiment, a two-class linear SVM on CLIP embeddings, drops from 96.9% video accuracy at full resolution to 72.8% at 25% resolution and 51.0% at 10% resolution, the latter at chance for balanced classes. Furthermore, Table 3 tests only the generic CLIP embedding and does not report corresponding results for the FT-CLIP embedding or the polynomial-kernel SVM that produce the headline 99.2% video accuracy in Table 1, so there is no evidence that the best-performing method is resolution-robust. Please add robustness results for the best-performing configuration or revise the abstract and contribution claims to specify that the robustness to resolution holds only for the CLIP linear-SVM variant and only down to 50% resolution.
[§4.2 and Table 2] The frame-to-prompt method is repeatedly described as requiring 'no explicit training' and as an unsupervised forensic technique, but its best performance in Table 2 is achieved with FT-CLIP, which was fine-tuned on the same DeepAction dataset using real/fake labels and captions like 'a REAL image' versus 'a FAKE image' (Appendix A.4). The 'no training' statement is therefore true only for the cosine-similarity decision rule, not for the underlying representation. Please state this explicitly, and report the frame-to-prompt results for the off-the-shelf CLIP, SigLIP, and JinaCLIP embeddings as the unsupervised baseline.
[§3 and §5] The DeepAction real class consists of 100 Pexels videos, while the fake class consists of 3,100 AI-generated clips, and the paper reports macro-averaged accuracy over five repetitions without any variance estimate. Because the video-level accuracies for the real class are based on only 100 videos, claims of >95% video accuracy should be accompanied by confidence intervals or per-repetition/per-class results. In addition, because the real videos are matched to the AI prompts only on coarse action category, differences in camera motion, lighting, and compression between Pexels footage and AI-generated clips may inflate the reported accuracies relative to in-the-wild performance; this limitation should be acknowledged in the generalization discussion.
[§5.3] The non-human-motion generalization experiment is missing a real-class control. Section 5.3 reports that the model 'correctly classified 97.1%' of 100 Sora/Runway videos as AI-generated, but all of these videos are AI-generated, so this number is a true-positive rate on the fake class, not an accuracy. Without a matched set of real non-human videos, the result cannot distinguish a genuinely semantic detector from a fake-biased classifier, and it does not support the claim that the model 'has not learned something specific to AI-generated human motion.' Please report a balanced accuracy or per-class recall/precision on a test set that includes real non-human videos from a comparable distribution.

minor comments (6)

[§5 (Table 2 discussion)] The sentence 'the P1 paired prompt authentic image vs. AI-generated image' appears to mislabel the prompt pair: P1 is 'real photo vs. fake photo,' while the authentic/AI-generated pair is P3. Please correct the cross-reference.
[§5 (Table 2 discussion)] The text states that the best frame-to-prompt result 'underperforms by only 6.6 percentage points' relative to 99.2%; the difference between 99.2% and the best frame-to-prompt video accuracy of 96.2% is 3.0 percentage points. Please correct this arithmetic.
[§5.1 (Table 3)] The table header describes compression as 'compression ratio measured as bits per second,' which conflates bitrate with compression ratio. Please use 'bitrate' and clarify that lower values correspond to stronger compression.
[§5.6] The comparison to prior work mixes metrics: AUC from [Vahdati et al.] is compared with accuracy from this paper. Please report the same metric or explicitly caution that the numbers are not directly comparable.
[Figures 3 and 4] The PCA projections in Figures 3 and 4 do not report the amount of variance explained by the displayed two principal components; please add this information so the separations can be interpreted quantitatively.
[Ethical Statement] The paper states that the code will not be open-sourced, which makes the reported preprocessing, fine-tuning, and SVM setups harder to reproduce; releasing the extraction and classification code would strengthen the reproducibility of the benchmark.

Circularity Check

1 steps flagged · score 6.0 of 10

Frame-to-prompt 'unsupervised' detection reduces to the supervised FT-CLIP fine-tuning objective, while the SVM evaluations are standard and non-circular.

fitted input called prediction [Section 4.2 (Frame-to-prompt) and Appendix A.4 (FT-CLIP), with results in Table 2.]
"In this third frame-to-prompt approach, each video-frame embedding is compared – through a simple cosine similarity – to an embedding of one of two prompts (e.g., “a real image” and “a fake image”). A frame is classified by selecting the class (real/fake) with the largest cosine similarity. ... The training set comprised two balanced sets of real and AI-generated frames, with captions constructed as “a {REAL/FAKE} image of {ACTION PROMPT}”."

FT-CLIP is fine-tuned on DeepAction with captions that literally encode the real/fake label ('a REAL image of {action}' vs 'a FAKE image of {action}'). The frame-to-prompt classifier then selects real/fake by cosine similarity to prompt pairs like 'a real image' vs 'a fake image'. After contrastive fine-tuning, this similarity score is exactly the training objective, so the high frame-to-prompt accuracy (95.2% with FT-CLIP vs 52.9% with base CLIP) is a readout of the supervised labels, not an unsupervised semantic property. The claim that 'frame-to-prompt approach requires no explicit training' is therefore misleading: the only variant that works was explicitly trained on the same dataset and label words. The prediction is forced by construction.

full rationale

The two-class and multi-class SVM pipelines (Table 1) are standard supervised evaluations on action-level held-out splits, and the leave-one-out generalization tests (Table 4) are independent of the fitted parameters; no circularity was found there. The robustness experiment (Table 3) undercuts the abstract's 'robust to resolution' claim but is an internal inconsistency, not a circular derivation. The significant circular step is the frame-to-prompt 'unsupervised' contribution: its best-performing variant uses FT-CLIP, which was fine-tuned on the same DeepAction dataset with captions of the form 'a REAL/FAKE image of {action}'. Because frame-to-prompt classifies by cosine similarity to 'a real image' vs 'a fake image', the classifier is the fine-tuning objective itself, and the near-chance performance of all untuned embeddings confirms that no unsupervised signal is being exploited. This makes the headline 'requires no explicit training' result reduce by construction to supervised training, warranting a partial-circularity score of 6.

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

The central claim rests on several domain assumptions about the dataset and embedding space, but not on any invented physical or mathematical entities. The free parameters are mostly training hyperparameters chosen by the authors.

free parameters (4)

FT-CLIP fine-tuning learning rates = 1e-5 (warm-up), 2e-3 (regular)
Reported in Appendix A.4; these hyperparameters affect the quality of the fine-tuned embedding and were chosen by hand, not derived.
FT-CLIP fine-tuning epochs = 1 warm-up + 1 regular
Same appendix; a training choice that impacts FT-CLIP performance.
FT-CLIP batch size = 16 training, 100 testing
Same appendix; chosen by authors.
SVM hyperparameters (C, gamma) = not reported
The paper says only implemented in scikit-learn and does not list kernel parameters; defaults likely used, but this is an unspecified modeling choice.

assumptions (4)

domain assumption Pre-trained multi-modal embeddings capture semantic features that separate AI-generated from real content.
The whole method assumes that CLIP-like embeddings encode meaningful real/fake distinctions (Section 4.1, Discussion). The paper admits it does not fully understand what properties drive the separation.
domain assumption Matching real and AI-generated videos only on action prompt removes semantic confounds.
Section 3 states the real videos were matched on general action to ensure embeddings are not based on semantic differences; this assumes stock footage from Pexels and AI clips are otherwise comparable.
domain assumption The prompt pair real photo versus fake photo aligns with actual real/fake status in the embedding space.
The frame-to-prompt method (Section 4.2) depends on this alignment; the near-chance accuracy of base CLIP and SigLIP in Table 2 shows this assumption is not generally true without fine-tuning.
domain assumption Leave-one-out performance over seven generators estimates generalization to unseen text-to-video models.
Section 5.2 uses leave-one-out to claim generalizability; the seven selected models are not a random sample of the space of generators.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Human Action CLIPs: Detecting AI-generated Human Motion." pith.science (2026). https://pith.science/paper/SU6PZTAJ

@misc{pith2026241200526,
  author       = {Pith},
  title        = {Pith review of: Human Action CLIPs: Detecting AI-generated Human Motion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SU6PZTAJ}},
  note         = {Machine review of arXiv:2412.00526}
}

read the original abstract

AI-generated video generation continues its journey through the uncanny valley to produce content that is increasingly perceptually indistinguishable from reality. To better protect individuals, organizations, and societies from its malicious applications, we describe an effective and robust technique for distinguishing real from AI-generated human motion using multi-modal semantic embeddings. Our method is robust to the types of laundering that typically confound more low- to mid-level approaches, including resolution and compression attacks. This method is evaluated against DeepAction, a custom-built, open-sourced dataset of video clips with human actions generated by seven text-to-video AI models and matching real footage. The dataset is available under an academic license at https://www.huggingface.co/datasets/faridlab/deepaction_v1.

Figures

Figures reproduced from arXiv: 2412.00526 by the authors.

**Figure 2.** Sample frames from one real and seven AI-generated videos prompted with “a person dancing to music”. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 4.** VideoPoet (blue) vs other models (orange) using [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

Survey on AI-Generated Media Detection: From Non-MLLM to MLLM
cs.CV 2025-02 unverdicted novelty 3.0 of 10

A survey organizing AI-generated media detection into Non-MLLM and MLLM based methods, with task and benchmark taxonomies.

Reference graph

Works this paper leans on

17 extracted references · 17 linked inside Pith · cited by 1 Pith paper

[3]

People are poorly equipped to detect AI-powered voice clones

Sarah Barrington and Hany Farid. People are poorly equipped to detect AI-powered voice clones. arXiv:2410.03791,

arXiv
[4]

Stable video diffusion: Scaling latent video diffusion models to large datasets

Andreas Blattmann, Tim Dockhorn, Sumith Kulal, Daniel Mendelevitch, Maciej Kilian, Dominik Lorenz, Yam Levi, Zion English, Vikram V oleti, Adam Letts, et al. Stable video diffusion: Scaling latent video diffusion models to large datasets. arXiv:2311.15127,

arXiv
[5]

Playing for 3D human recovery

Zhongang Cai, Mingyuan Zhang, Jiawei Ren, Chen Wei, Daxuan Ren, Zhengyu Lin, Haiyu Zhao, Lei Yang, Chen Change Loy, and Ziwei Liu. Playing for 3D human recovery. arXiv:2110.07588,

arXiv
[6]

PaLI: A jointly-scaled multilingual language-image model

Xi Chen, Xiao Wang, Soravit Changpinyo, AJ Piergio- vanni, Piotr Padlewski, Daniel Salz, Sebastian Good- man, Adam Grycner, Basil Mustafa, Lucas Beyer, et al. PaLI: A jointly-scaled multilingual language-image model. arXiv:2209.06794,

arXiv
[7]

Expos- ing lip-syncing deepfakes from mouth inconsistencies

Soumyya Kanti Datta, Shan Jia, and Siwei Lyu. Expos- ing lip-syncing deepfakes from mouth inconsistencies. arXiv:2401.10113,

arXiv
[8]

Exploring the adver- sarial robustness of CLIP for AI-generated image detec- tion

Vincenzo De Rosa, Fabrizio Guillaro, Giovanni Poggi, Da- vide Cozzolino, and Luisa Verdoliva. Exploring the adver- sarial robustness of CLIP for AI-generated image detec- tion. arXiv:2407.19553,

arXiv
[9]

VideoPoet: A large language model for zero-shot video generation

Dan Kondratyuk, Lijun Yu, Xiuye Gu, Jos´e Lezama, Jonathan Huang, Rachel Hornung, Hartwig Adam, Hassan Ak- bari, Yair Alon, Vighnesh Birodkar, et al. VideoPoet: A large language model for zero-shot video generation. arXiv:2312.14125,

arXiv
[11]

AnimateDiff-Lightning: Cross-model diffusion distillation

Shanchuan Lin and Xiao Yang. AnimateDiff-Lightning: Cross-model diffusion distillation. arXiv:2403.12706,

arXiv

Show all 17 references

[14]

DeCLIP: Decoding CLIP representations for deepfake localization

Stefan Smeu, Elisabeta Oneata, and Dan Oneata. DeCLIP: Decoding CLIP representations for deepfake localization. arXiv:2409.08849,

arXiv
[15]

CLIP models are few-shot learners: Empirical studies on vqa and visual entailment

Haoyu Song, Li Dong, Wei-Nan Zhang, Ting Liu, and Furu Wei. CLIP models are few-shot learners: Empirical studies on vqa and visual entailment. arXiv:2203.07190,

arXiv
[17]

4Real: Towards photo- realistic 4D scene generation via video diffusion models

Heng Yu, Chaoyang Wang, Peiye Zhuang, Willi Menapace, Aliaksandr Siarohin, Junli Cao, Laszlo A Jeni, Sergey Tulyakov, and Hsin-Ying Lee. 4Real: Towards photo- realistic 4D scene generation via video diffusion models. arXiv:2406.07472,

arXiv
[2001]

CogVideoX: Text- to-video diffusion models with an expert transformer

Zhuoyi Yang, Jiayan Teng, Wendi Zheng, Ming Ding, Shiyu Huang, Jiazheng Xu, Yuanming Yang, Wenyi Hong, Xi- aohan Zhang, Guanyu Feng, et al. CogVideoX: Text- to-video diffusion models with an expert transformer. arXiv:2408.06072,

arXiv
[2020]

Vidu: A highly consistent, dynamic and skilled text-to-video generator with diffusion models

Fan Bao, Chendong Xiang, Gang Yue, Guande He, Hongzhou Zhu, Kaiwen Zheng, Min Zhao, Shilong Liu, Yaole Wang, and Jun Zhu. Vidu: A highly consistent, dynamic and skilled text-to-video generator with diffusion models. arXiv:2405.04233,

arXiv
[2021]

LAION- 400M: Open dataset of clip-filtered 400 million image-text pairs

Christoph Schuhmann, Richard Vencu, Romain Beaumont, Robert Kaczmarczyk, Clayton Mullis, Aarush Katta, Theo Coombes, Jenia Jitsev, and Aran Komatsuzaki. LAION- 400M: Open dataset of clip-filtered 400 million image-text pairs. arXiv:2111.02114,

arXiv
[2022]

How much can CLIP benefit vision-and-language tasks? arXiv:2107.06383,

Sheng Shen, Liunian Harold Li, Hao Tan, Mohit Bansal, Anna Rohrbach, Kai-Wei Chang, Zhewei Yao, and Kurt Keutzer. How much can CLIP benefit vision-and-language tasks? arXiv:2107.06383,

arXiv
[2023]

Jina CLIP: Your CLIP model is also your text retriever

Andreas Koukounas, Georgios Mastrapas, Michael G ¨unther, Bo Wang, Scott Martens, Isabelle Mohr, Saba Sturua, Mo- hammad Kalim Akram, Joan Fontanals Mart ´ınez, Saahil Ognawala, et al. Jina CLIP: Your CLIP model is also your text retriever. arXiv:2405.20204,

arXiv
[2024]

Lu- miere: A space-time diffusion model for video generation

Omer Bar-Tal, Hila Chefer, Omer Tov, Charles Her- rmann, Roni Paiss, Shiran Zada, Ariel Ephrat, Jun- hwa Hur, Yuanzhen Li, Tomer Michaeli, et al. Lu- miere: A space-time diffusion model for video generation. arXiv:2401.12945,

arXiv

Pith tools

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

[1] [3]

People are poorly equipped to detect AI-powered voice clones

Sarah Barrington and Hany Farid. People are poorly equipped to detect AI-powered voice clones. arXiv:2410.03791,

arXiv

[2] [4]

Stable video diffusion: Scaling latent video diffusion models to large datasets

Andreas Blattmann, Tim Dockhorn, Sumith Kulal, Daniel Mendelevitch, Maciej Kilian, Dominik Lorenz, Yam Levi, Zion English, Vikram V oleti, Adam Letts, et al. Stable video diffusion: Scaling latent video diffusion models to large datasets. arXiv:2311.15127,

arXiv

[3] [5]

Playing for 3D human recovery

Zhongang Cai, Mingyuan Zhang, Jiawei Ren, Chen Wei, Daxuan Ren, Zhengyu Lin, Haiyu Zhao, Lei Yang, Chen Change Loy, and Ziwei Liu. Playing for 3D human recovery. arXiv:2110.07588,

arXiv

[4] [6]

PaLI: A jointly-scaled multilingual language-image model

Xi Chen, Xiao Wang, Soravit Changpinyo, AJ Piergio- vanni, Piotr Padlewski, Daniel Salz, Sebastian Good- man, Adam Grycner, Basil Mustafa, Lucas Beyer, et al. PaLI: A jointly-scaled multilingual language-image model. arXiv:2209.06794,

arXiv

[5] [7]

Expos- ing lip-syncing deepfakes from mouth inconsistencies

Soumyya Kanti Datta, Shan Jia, and Siwei Lyu. Expos- ing lip-syncing deepfakes from mouth inconsistencies. arXiv:2401.10113,

arXiv

[6] [8]

Exploring the adver- sarial robustness of CLIP for AI-generated image detec- tion

Vincenzo De Rosa, Fabrizio Guillaro, Giovanni Poggi, Da- vide Cozzolino, and Luisa Verdoliva. Exploring the adver- sarial robustness of CLIP for AI-generated image detec- tion. arXiv:2407.19553,

arXiv

[7] [9]

VideoPoet: A large language model for zero-shot video generation

Dan Kondratyuk, Lijun Yu, Xiuye Gu, Jos´e Lezama, Jonathan Huang, Rachel Hornung, Hartwig Adam, Hassan Ak- bari, Yair Alon, Vighnesh Birodkar, et al. VideoPoet: A large language model for zero-shot video generation. arXiv:2312.14125,

arXiv

[8] [11]

AnimateDiff-Lightning: Cross-model diffusion distillation

Shanchuan Lin and Xiao Yang. AnimateDiff-Lightning: Cross-model diffusion distillation. arXiv:2403.12706,

arXiv