Zero-Shot Personalized Camera Motion Control for Image-to-Video Synthesis
arXiv:2504.09472v3 Announce Type: replace Abstract: Specifying nuanced and compelling camera motion remains a significant hurdle for non-expert creators using generative tools, creating an "expressive gap" where generic text prompts fail to capture cinematic vision. This barrier limits individual creativity and restricts the accessibility of cinematic production for small-scale industries and educational content creators. To address this, we present a zero-shot diffusion-based framework for personalized camera motion control, enabling the transfer of cinematic movements from a single reference — Pooja Guhan, Divya Kothandaraman, Geonsun Lee, Tsung-Wei Huang, Guan-Ming Su, Dinesh Manocha
View PDF HTML (experimental)
Abstract:Specifying nuanced and compelling camera motion remains a significant hurdle for non-expert creators using generative tools, creating an "expressive gap" where generic text prompts fail to capture cinematic vision. This barrier limits individual creativity and restricts the accessibility of cinematic production for small-scale industries and educational content creators. To address this, we present a zero-shot diffusion-based framework for personalized camera motion control, enabling the transfer of cinematic movements from a single reference video onto a user-provided static image without requiring 3D data, predefined trajectories, or complex graphical interfaces. Our technical contribution involves an inference-time optimization strategy using dual Low-Rank Adaptation (LoRA) networks, with an orthogonality regularizer that encourages separation between spatial appearance and temporal motion updates, alongside a homography-based refinement strategy that provides weak geometric guidance. We evaluate our approach using a new metric, CameraScore, and two distinct user studies. A 72-participant perceptual study demonstrates that our method significantly outperforms existing baselines in motion accuracy (90.45% preference) and scene preservation (70.31% preference). Furthermore, a 12-participant task-based interaction study confirms that our workflow significantly improves usability and creative control (p < 0.001) compared to standard text- or preset-based prompts. We hope this work lays a foundation for future advancements in camera motion transfer across diverse scenes.
Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Cite as: arXiv:2504.09472 [cs.CV]
(or arXiv:2504.09472v3 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2504.09472
arXiv-issued DOI via DataCite
Submission history
From: Pooja Guhan [view email] [v1] Sun, 13 Apr 2025 08:04:11 UTC (18,777 KB) [v2] Tue, 23 Dec 2025 04:09:07 UTC (4,367 KB) [v3] Fri, 27 Mar 2026 06:33:38 UTC (19,030 KB)
Sign in to highlight and annotate this article
Conversation starters
Daily AI Digest
Get the top 5 AI stories delivered to your inbox every morning.
More about
researchpaperarxiv
Assessing Pause Thresholds for empirical Translation Process Research
arXiv:2604.01410v1 Announce Type: new Abstract: Text production (and translations) proceeds in the form of stretches of typing, interrupted by keystroke pauses. It is often assumed that fast typing reflects unchallenged/automated translation production while long(er) typing pauses are indicative of translation problems, hurdles or difficulties. Building on a long discussion concerning the determination of pause thresholds that separate automated from presumably reflective translation processes (O'Brien, 2006; Alves and Vale, 2009; Timarova et al., 2011; Dragsted and Carl, 2013; Lacruz et al., 2014; Kumpulainen, 2015; Heilmann and Neumann 2016), this paper compares three recent approaches for computing these pause thresholds, and suggest and evaluate a novel method for computing Production
Friends and Grandmothers in Silico: Localizing Entity Cells in Language Models
arXiv:2604.01404v1 Announce Type: new Abstract: Language models can answer many entity-centric factual questions, but it remains unclear which internal mechanisms are involved in this process. We study this question across multiple language models. We localize entity-selective MLP neurons using templated prompts about each entity, and then validate them with causal interventions on PopQA-based QA examples. On a curated set of 200 entities drawn from PopQA, localized neurons concentrate in early layers. Negative ablation produces entity-specific amnesia, while controlled injection at a placeholder token improves answer retrieval relative to mean-entity and wrong-cell controls. For many entities, activating a single localized neuron is sufficient to recover entity-consistent predictions once
Knowledge Map
Connected Articles — Knowledge Graph
This article is connected to other articles through shared AI topics and tags.
More in Research Papers
Assessing Pause Thresholds for empirical Translation Process Research
arXiv:2604.01410v1 Announce Type: new Abstract: Text production (and translations) proceeds in the form of stretches of typing, interrupted by keystroke pauses. It is often assumed that fast typing reflects unchallenged/automated translation production while long(er) typing pauses are indicative of translation problems, hurdles or difficulties. Building on a long discussion concerning the determination of pause thresholds that separate automated from presumably reflective translation processes (O'Brien, 2006; Alves and Vale, 2009; Timarova et al., 2011; Dragsted and Carl, 2013; Lacruz et al., 2014; Kumpulainen, 2015; Heilmann and Neumann 2016), this paper compares three recent approaches for computing these pause thresholds, and suggest and evaluate a novel method for computing Production
![[R], 31 MILLIONS High frequency data, Light GBM worked perfectly](https://d2xsxph8kpxj0f.cloudfront.net/310419663032563854/konzwo8nGf8Z4uZsMefwMr/default-img-neural-network-P6fqXULWLNUwjuxqUZnB3T.webp)
[R], 31 MILLIONS High frequency data, Light GBM worked perfectly
We just published a paper on predicting adverse selection in high-frequency crypto markets using LightGBM , and I wanted to share it here because the findings are directly relevant to anyone dealing high frequency data and machine learning The core problem we solved: Every market maker's nightmare — getting picked off by informed traders right before a big move. We built a model that flags those toxic seconds before they wreck you. The data: - 31,081,463 second-level observations of BTC/USDT perpetual futures on Bybit - February 2025 → February 2026 (381 raw daily files) - Strict walk-forward regime, zero lookahead bias The key results (this is the part that shocked us): Our TailScore metric — which combines predicted toxicity probability with predicted price move severity — flags the top
![[D] ACL 2026 Decision](https://d2xsxph8kpxj0f.cloudfront.net/310419663032563854/konzwo8nGf8Z4uZsMefwMr/default-img-satellite-bUaXYHZsoMZjyA4XgfFqkD.webp)
Discussion
Sign in to join the discussion
No comments yet — be the first to share your thoughts!