Camera-Controlled Video Diffusion

TL;DR

Video diffusion models conditioned on explicit camera trajectories (and increasingly on object trajectories too) to re-render or extend a scene under a user-specified viewpoint path. The dominant recent recipe in this cluster is a frozen Wan2.1 latent video diffusion backbone with a lightweight geometric adapter injecting camera (and optionally object) conditioning — ReCamMaster-style camera-embedder lineage. The three filed papers triangulate the design space along a control axis: pure camera (ReCamMaster, ViewCrafter, FlashWorld, Voyager families), space+time (SpaceTimePilot), and space+multi-object (VerseCrafter); NeoVerse threads camera control through a feed-forward 4DGS reconstructor.

Key claims

• “Frozen large video diffusion backbone + small trainable geometric adapter” is the de-facto recipe across the filed papers — SpaceTimePilot freezes Wan2.1 T2V-1.3B and trains only E_cam, E_ani, self-attention, and a projection layer (SpaceTimePilot: Generative Rendering of Dynamic Scenes Across Space and Time §3); NeoVerse freezes Wan-T2V-14B and trains only a control branch (NeoVerse: Enhancing 4D World Model with in-the-wild Monocular Videos §3.2); VerseCrafter freezes Wan2.1-14B and trains only the GeoAdapter blocks attached every 5th DiT layer (VerseCrafter: Dynamic Realistic Video World Model with 4D Geometric Control §3.2); OmniTransfer extends this to a unified reference-branch adapter on Wan2.1-I2V-14B that handles camera alongside ID/style/motion/effect (OmniTransfer: All-in-one Framework for Spatio-temporal Video Transfer §4); Wan-Animate post-trains Wan-I2V for character animation/replacement with a region-flagged input paradigm and skeleton+expression decoupling, without per-block adapters (Wan-Animate: Unified Character Animation and Replacement with Holistic Replication §3).
• Source-aware camera conditioning — jointly encoding source and target trajectories rather than assuming a shared first frame — fixes ReCamMaster’s first-frame-copy artifact and is necessary for full trajectory control (SpaceTimePilot: Generative Rendering of Dynamic Scenes Across Space and Time §3.3, Eq. 3).
• Disentangling camera from object/animation control requires separate conditioning channels, not a shared one: SpaceTimePilot adds an animation-time embedder E_ani independent of the RoPE frame-index (SpaceTimePilot: Generative Rendering of Dynamic Scenes Across Space and Time §3.2.1, Eq. 2); VerseCrafter renders background and 3D Gaussian trajectory maps through decoupled channels and ablates the merged variant to show ObjMC degrades from 2.51 to 3.73 (VerseCrafter: Dynamic Realistic Video World Model with 4D Geometric Control Table 3, §5.3); OmniTransfer disentangles appearance vs temporal tasks via the axis of the RoPE offset alone — spatial-width offset for temporal/camera tasks, temporal-frame offset for ID/style (OmniTransfer: All-in-one Framework for Spatio-temporal Video Transfer §4.2, Eq. 2); Wan-Animate decouples body-motion (skeleton added to noise latents) from expression (implicit features through standard conditioning paths) (Wan-Animate: Unified Character Animation and Replacement with Holistic Replication §3); Kling-MotionControl goes further to three region-tailored motion channels (body / face / hand) plus a fourth explicit semantic-motion channel on top of the geometric one (Kling-MotionControl Technical Report §1).
• Object motion is best represented as 3D Gaussian trajectories rather than 3D bounding boxes or 3D point trajectories: ObjMC 2.51 (Gaussian) vs 4.52 (box) vs 6.90 (point), with VBench overall 88.10 vs 85.45 vs 85.57 (VerseCrafter: Dynamic Realistic Video World Model with 4D Geometric Control Table 3).
• Depth channels matter — both SpaceTimePilot’s source-aware encoding and VerseCrafter’s RGB+depth control maps are reported as load-bearing; removing depth from VerseCrafter’s control raises RotErr from 0.89 to 1.18 and ObjMC from 2.51 to 4.93 (VerseCrafter: Dynamic Realistic Video World Model with 4D Geometric Control Table 3, §5.3).
• Reconstruction-based camera control trades visual quality for trajectory accuracy vs. purely generative camera control — NeoVerse explicitly observes that ReCamMaster has higher VBench quality but weaker trajectory control while TrajectoryCrafter has the opposite, and claims its hybrid 4DGS+diffusion design pushes out both axes (NeoVerse: Enhancing 4D World Model with in-the-wild Monocular Videos §4.3, Fig. 4).
• The data bottleneck for joint camera+object control is structural: real captures cannot replay a dynamic scene at different temporal/camera-trajectory mappings, so all three filed papers manufacture supervision synthetically — CamTime/temporal-warping (SpaceTimePilot: Generative Rendering of Dynamic Scenes Across Space and Time), online geometric-degradation simulation (NeoVerse: Enhancing 4D World Model with in-the-wild Monocular Videos), automatic 4D Gaussian-trajectory annotation on monocular video (VerseCrafter: Dynamic Realistic Video World Model with 4D Geometric Control §4). OmniTransfer similarly resorts to internally collected reference-pair data, citing the absence of any public dataset (OmniTransfer: All-in-one Framework for Spatio-temporal Video Transfer §5.1). Kling-MotionControl augments with high-quality rendering data and high-speed-camera footage to specifically support rapid/complex motion (Kling-MotionControl Technical Report §1).
• “Multimodal guidance” (dense + sparse control signals fused into the conditioning stream) is being framed as an implicit world-level supervision signal at the multi-minute AR generation regime, distinct from the explicit-camera-trajectory recipes above (LongVie 2: Multimodal Controllable Ultra-Long Video World Model project page).
• Reference-video conditioning is a viable alternative to explicit trajectory parameterization for camera control: OmniTransfer beats both MotionClone (attention-inversion) and CamCloneMaster (temporal context concatenation) by ~2.5× on user-study camera fidelity (4.19 vs 1.79) without any explicit trajectory input, using only a clean reference-video latent injected via a RoPE-offset reference branch (OmniTransfer: All-in-one Framework for Spatio-temporal Video Transfer Table 5, §5.3.4).
• One-directional (causal) reference→target attention prevents copy-paste artifacts that bidirectional joint self-attention introduces, and cutting the reference branch’s diffusion-time embedding to a fixed t=0 makes the reference encoder time-invariant — together yielding a 20% inference-time reduction on Wan2.1-14B (OmniTransfer: All-in-one Framework for Spatio-temporal Video Transfer §4.3, Table 7).
• Spatially-aligned control signals can be injected by direct addition to noise latents, with no learned adapter on that path — Wan-Animate adds a 2D skeleton image to the initial noise latents for body-motion control and only uses the existing conditioning pathway for the non-aligned expression signal (Wan-Animate: Unified Character Animation and Replacement with Holistic Replication §3); contrasts with VerseCrafter’s GeoAdapter approach (which is also spatially aligned but uses learned adapter blocks every 5th DiT layer).
• Character motion and camera trajectory control can be composed on the same backbone: Kling-MotionControl endows its multi-granular motion representations with 3D awareness via large-scale multi-view supervision, enabling text-driven cinematic camera control (pans, zooms, free-view) on top of body/face/hand motion in a single model (Kling-MotionControl Technical Report §1) — first filed concrete instance of the “stack adapters / channels for camera + character” open question.
• Multi-stage CFG-merging distillation on top of holistic-control DiTs gives >10× inference speedup while preserving GSB-evaluated quality, combining a dual-branch teacher sampling for multi-conditional CFG, NFE reduction via distillation, and conditional-gradient merging into the student (Kling-MotionControl Technical Report §1, §2.2).
• In closed-vs-closed character animation, Kling-MotionControl is preferred over Wan-Animate, Dreamina, and Runway Act-Two on all five granular GSB dimensions (visual / dynamic / ID / motion / expression) plus Overall (Kling-MotionControl Technical Report Table 1, §2.2). For Wan-Animate specifically the Overall (G+S)/(B+S) ratio is 4.00 with Expression 1.16 and ID 3.07.
• Temporal RoPE warping can be driven by a measured property of the model’s own output, not just by an external control signal: SPF/ReTime (Video Analysis and Generation via a Semantic Progress Function) fits a Semantic Progress Function to SigLIP frame-distance embeddings, then inverts it to warp temporal RoPE positions toward uniform semantic velocity — with frequency-aware blending (stronger on low-frequency RoPE bands), a timestep-decay schedule, and three refinement iterations. Same RoPE-manipulation primitive as SpaceTimePilot’s E_ani and OmniTransfer’s offset axes, but the conditioning signal is the SPF itself (Video Analysis and Generation via a Semantic Progress Function §4.1, Eq. 7–11). Generalizes to any target pacing function (rising/falling exponentials), and the audio in LTX-2 naturally realigns because cross-modal attention is anchored to linear temporal coordinates.

Recent contributions

• [2026-07-14] Motion4Motion: Motion Transfer Across Subjects at Inference: Motion4Motion (SIGGRAPH ‘26) is a training-free, skeleton-free motion-transfer recipe on frozen WAN-T2V: extract per-anchor motion flow from a source (Grounded SAM-2 + diffusion-feature matching + point tracking), then inject it into self-attention by caching target K/V and re-embedding them with new RoPE positions from the flow (“TransPE”). Third filed instance of inference-time RoPE manipulation as the control primitive on WAN (after ReTime and OmniTransfer), and the first to cover the cross-topology regime (human → goose → panda → animate table) that skeleton-based methods like Wan-Animate and Kling-MotionControl cannot bridge.
• [2026-06-19] MolmoMotion: Language-guided 3D motion forecasting: MolmoMotion’s predicted 3D point trajectories are used as the motion-control signal to DaS on top of CogVideoX-5B; the combination is reported to beat the much larger Wan2.2-I2V-A14B on four of five VBench motion metrics — a concrete recipe for trajectory-conditioning I2V via a separately pretrained language-conditioned motion forecaster rather than baking trajectory control into the diffusion model.
• [2026-06-09] Latent Spatial Memory for Video World Models: Mirage realizes camera control implicitly: novel-view synthesis is direct latent warping of cached tokens into the target view via depth-guided projection, with no explicit camera-trajectory conditioning channel. Reports state-of-the-art WorldScore (70.36 avg) and strong RealEstate10K reconstruction. Sibling recipe to point-cloud-rendering-as-ControlNet (SPMem, NeoVerse), but indexing latents instead of RGB.
• [2026-06-02] EgoX: Egocentric Video Generation from a Single Exocentric Video: EgoX (KAIST AI / DAVIAN Robotics) pushes the “frozen video DiT + LoRA + decoupled conditioning” recipe to the extreme no-overlap regime — exo-to-ego viewpoint translation. Uses 4DNeX as the frozen backbone, LoRA adaptation, a unified conditioning scheme (width-wise concat of the exocentric source video + channel-wise concat of an estimated egocentric prior video), and geometry-guided self-attention that restricts attention to spatially relevant regions. Trained on EgoExo4D, reports SOTA on exo-to-ego benchmarks and generalizes to in-the-wild clips (Dark Knight). First filed paper in the cluster that explicitly targets the zero-pixel-overlap camera transformation case; the egocentric prior plays the structural role the depth / Gaussian-trajectory channels play in VerseCrafter and SpaceTimePilot.
• [2026-06-02] PoseAnything: Universal Pose-guided Video Generation with Part-aware Temporal Coherence: PoseAnything (Wan2.2-TI2V-5B base) introduces Subject and Camera Motion Decoupled CFG — subject-pose conditioning to the positive CFG anchor, camera-motion conditioning to the negative anchor — yielding camera-trajectory control at inference time from a model trained only on subject-pose data. First filed instance of using CFG anchor structure (not a trained control channel) as the camera-control primitive. Also ships XPose, a 50K-clip non-human pose-video dataset for universal-skeleton pose-driven generation.
• [2026-05-24] Video Analysis and Generation via a Semantic Progress Function: Introduces a measure-then-warp paradigm for temporal control on Wan2.2 and LTX-2 — fits a 1D Semantic Progress Function to SigLIP pairwise embedding distances, then inverts it to warp temporal RoPE positions toward uniform semantic velocity. Frequency-aware blending of warping strength across RoPE bands, exponential timestep-decay multiplier, three iterations. Within 1σ of original on VBench Aesthetic / Motion / Temporal Flicker, 88% user-study preference on pacing. For closed-source first-last-frame models, segments the SPF by piecewise-linear least squares and regenerates clips per segment. LTX-2 audio automatically realigns because the model anchors cross-modal attention to linear temporal coordinates. SIGGRAPH ‘26.
• [2026-05-23] Kling-MotionControl Technical Report: Kuaishou’s Kling-MotionControl — unified DiT for holistic character animation with three-channel divide-and-conquer (body/face/hand) motion + an explicit semantic-motion track, identity-agnostic geometric motion learning, a subject-library multi-reference mechanism, 3D-aware motion representations enabling text-driven cinematic camera control composed with character motion, and a multi-stage CFG-merged distillation stack yielding >10× inference speedup. GSB human evaluation reports preference over Wan-Animate (Overall 4.00), Dreamina (3.44), and Runway Act-Two (16.25) on a 150-case 1080p benchmark across all five granular dimensions.
• [2026-05-23] Wan-Animate: Unified Character Animation and Replacement with Holistic Replication: Wan-Animate post-trains Wan-I2V (productized as Wan 2.2 Animate) for unified character animation + character replacement. Uses a region-flagged input paradigm that preserves Wan-I2V’s latent-input structure, decouples body-motion (spatially-aligned 2D skeleton added to noise latents) from expression (implicit features), and adds a Relighting LoRA only for replacement mode. Reports SOTA against Animate Anyone / UniAnimate / VACE on SSIM/LPIPS/FVD and user-study preference over DreamActor-M1 and Runway Act-Two; productized May 2026 on fal as “Wan Motion” with built-in motion retargeting and optional Flux-2 Edit identity sharpening.
• [2026-05-22] OmniTransfer: All-in-one Framework for Spatio-temporal Video Transfer: Unifies camera transfer with ID/style/motion/effect transfer in a single Wan2.1-I2V-14B model. Replaces explicit camera-trajectory parameterization with a clean reference-video branch and a per-task RoPE offset (spatial-width for temporal tasks, temporal-frame for appearance tasks). Adds Reference-decoupled Causal Learning (one-way reference→target attention + fixed t=0 on reference branch → 20% faster inference) and a Qwen-2.5-VL-with-MetaQueries connector replacing T5. Beats MotionClone and CamCloneMaster ~2.5× on user-study camera fidelity. Argues this is the first cross-task reference-conditioning recipe that generalizes to unseen task combinations.
• [2026-05-22] LongVie 2: Multimodal Controllable Ultra-Long Video World Model: Stretches the cluster from clip-scale (5–10s) camera/object control to multi-minute AR generation. Treats control as one of three sequenced training stages (multimodal guidance → degradation-aware training → history-context guidance) rather than a single conditioning channel. Project-page-only; specifics of which control signals (camera trajectories? sparse poses? depth maps?) not yet disclosed.
• [2026-05-22] VerseCrafter: Dynamic Realistic Video World Model with 4D Geometric Control: Introduces 3D Gaussian trajectories as the object-motion control space, with a decoupled background + foreground GeoAdapter on frozen Wan2.1-14B; ablates against 3D bounding boxes, 3D point trajectories, depth removal, and merged-channel variants; builds VerseControl4D (35k clips) with fully automatic 4D annotation.
• [2026-05-22] NeoVerse: Enhancing 4D World Model with in-the-wild Monocular Videos: Threads camera control through a feed-forward 4DGS reconstructor (“Gaussianized” VGGT) + frozen Wan-T2V-14B; introduces a quality/compute knob (11→full keyframes) and achieves 6–8× faster inference than TrajectoryCrafter and ReCamMaster at parity or better VBench.
• [2026-05-22] SpaceTimePilot: Generative Rendering of Dynamic Scenes Across Space and Time: Disentangles space (camera) and time (animation) control on a frozen Wan-2.1 T2V-1.3B with an animation-time embedder and source-aware camera conditioning; supplies the synthetic CamTime dataset (180k Blender videos, full camera×time grid).

Open questions

• Does a single frozen Wan + multiple specialized adapters (camera, time, object, character) compose? The filed papers all train one adapter at a time on top of Wan; OmniTransfer composes 5 task switches inside a single adapter but does not stack independent adapters — nobody has demonstrated that VerseCrafter’s GeoAdapter, SpaceTimePilot’s E_ani, and Wan-Animate’s character/skeleton head can be combined. Kling-MotionControl is the first instance of camera + character control inside one model (via 3D-aware motion representations), but it is a from-scratch DiT and does not test composability with the Wan-adapter recipes. ReTime is composable in principle — it warps RoPE at inference time and doesn’t touch weights — but composability with adapter-trained warping has not been tested.
• Where does 3D Gaussian trajectory representation break? VerseCrafter notes ellipsoid ambiguity for near-isotropic shapes (humans) — under what data distribution does this hurt enough to justify hybrid Gaussian + SMPL-X? Wan-Animate’s skeleton-based body-motion channel and Kling-MotionControl’s three-channel body/face/hand decomposition are two concrete alternatives for the human-character case.
• Reconstruction-based (NeoVerse) vs control-first (VerseCrafter, SpaceTimePilot, Wan-Animate, Kling-MotionControl) vs reference-video-conditioning (OmniTransfer) vs measure-then-warp (ReTime): now four-plus quite different paradigms. Is there a regime where one dominates? OmniTransfer beats CamCloneMaster on user-study camera fidelity but has not been head-to-head with ReCamMaster / NeoVerse / Kling-MotionControl on VBench-style trajectory metrics.
• The control-axis explosion (camera, time, multi-object, character, lighting, semantic pacing?) implies a combinatorial trajectory-annotation cost. OmniTransfer sidesteps part of this by using reference videos themselves as the control signal; Kling-MotionControl sidesteps it for character-motion via identity-agnostic geometric motion learning + semantic motion tagging; ReTime sidesteps it entirely by deriving the control signal from the model’s own output via SPF.
• All five Wan-based papers use Wan2.x as the frozen prior (Wan-Animate uses 2.2; the others 2.1; ReTime uses 2.2). How much of the reported quality comes from Wan itself vs. the geometric-adapter / reference-branch / region-flag design choices? Kling-MotionControl is the contrast point — it does not start from Wan and still reports preference over Wan-Animate.
• Does camera/object control survive the autoregressive multi-minute regime? LongVie 2 claims controllability at the 5-minute horizon, but no head-to-head with clip-scale ReCamMaster / VerseCrafter / OmniTransfer / Wan-Animate / Kling-MotionControl quality has been reported.
• Is OmniTransfer’s “fixed t=0 on reference branch → time-invariant encoder → free 20% inference” trick transferable to the explicit-trajectory recipes? They all run the geometric adapter at every diffusion step today. Kling-MotionControl’s CFG-merging + multi-stage distillation gets to >10× speedup via a different route (student NFE reduction); the two acceleration tricks are likely composable. ReTime’s timestep-decay multiplier β(t) is a third complementary acceleration heuristic — concentrates work at early diffusion steps.
• Wan-Animate gets away with additive injection of a spatially-aligned skeleton map onto noise latents — no adapter on that path. Could the explicit-camera-trajectory recipes (which use spatially-aligned RGB+depth control maps) drop their adapter blocks in favor of an additive injection too?
• Three-channel (Kling-MotionControl, body/face/hand) vs two-channel (Wan-Animate, body/face) character-motion decomposition: is the third hand channel load-bearing, or is the win attributable to the explicit semantic-motion track, the multi-stage training, or the 3D-aware representation? No ablation table separates these in Kling-MotionControl’s report.
• ReTime warps temporal RoPE uniformly across the whole clip toward linear semantic velocity. Could a per-region warp (e.g., enforced only on a subject mask) compose with the global trajectory adapters above for region-specific pacing control?

Papers

2026-05

• LongVie 2: Multimodal Controllable Ultra-Long Video World Model — LongVie 2: Multimodal Controllable Ultra-Long Video World Model (published 2026-05-22)
• SANA-WM: Efficient Minute-Scale World Modeling with Hybrid Linear Diffusion Transformer — SANA-WM: Efficient Minute-Scale World Modeling with Hybrid Linear Diffusion Transformer (published 2026-05-14)

2026-04

• Video Analysis and Generation via a Semantic Progress Function — Video Analysis and Generation via a Semantic Progress Function (published 2026-04-24)
• Reshoot-Anything: A Self-Supervised Model for In-the-Wild Video Reshooting — Reshoot-Anything: A Self-Supervised Model for In-the-Wild Video Reshooting (published 2026-04-23)
• Vista4D: Video Reshooting with 4D Point Clouds — Vista4D: Video Reshooting with 4D Point Clouds (published 2026-04-23)
• Lyra 2.0: Explorable Generative 3D Worlds — Lyra 2.0: Explorable Generative 3D Worlds (published 2026-04-15)
• MoRight: Motion Control Done Right — MoRight: Motion Control Done Right (published 2026-04-10)
• CHAI: Building a Precise Video Language with Human–AI Oversight — CHAI: Building a Precise Video Language with Human–AI Oversight (published 2026-04-01)

2026-03

• Kling-MotionControl Technical Report — Kling-MotionControl Technical Report (published 2026-03-03)

2026-01

• Advancing Open-source World Models (LingBot-World) — Advancing Open-source World Models (LingBot-World) (published 2026-01-29)
• 360Anything: Geometry-Free Lifting of Images and Videos to 360° — 360Anything: Geometry-Free Lifting of Images and Videos to 360° (published 2026-01-28)
• OmniTransfer: All-in-one Framework for Spatio-temporal Video Transfer — OmniTransfer: All-in-one Framework for Spatio-temporal Video Transfer (published 2026-01-20)
• VerseCrafter: Dynamic Realistic Video World Model with 4D Geometric Control — VerseCrafter: Dynamic Realistic Video World Model with 4D Geometric Control (published 2026-01-08)
• NeoVerse: Enhancing 4D World Model with in-the-wild Monocular Videos — NeoVerse: Enhancing 4D World Model with in-the-wild Monocular Videos (published 2026-01-01)

2025-12

• SpaceTimePilot: Generative Rendering of Dynamic Scenes Across Space and Time — SpaceTimePilot: Generative Rendering of Dynamic Scenes Across Space and Time (published 2025-12-31)
• InfCam: Infinite-Homography as Robust Conditioning for Camera-Controlled Video Generation — InfCam: Infinite-Homography as Robust Conditioning for Camera-Controlled Video Generation (published 2025-12-22)
• HY-World 1.5 (WorldPlay): A Systematic Framework for Interactive World Modeling with Real-Time Latency and Geometric Consistency — HY-World 1.5 (WorldPlay): A Systematic Framework for Interactive World Modeling with Real-Time Latency and Geometric Consistency (published 2025-12-17)
• Wan-Move: Motion-controllable Video Generation via Latent Trajectory Guidance — Wan-Move: Motion-controllable Video Generation via Latent Trajectory Guidance (published 2025-12-09)
• Generative Video Motion Editing with 3D Point Tracks — Generative Video Motion Editing with 3D Point Tracks (published 2025-12-01)

2025-11

• Captain Safari: A World Engine with Pose-Aligned 3D Memory — Captain Safari: A World Engine with Pose-Aligned 3D Memory (published 2025-11-28)
• SteadyDancer: Harmonized and Coherent Human Image Animation with First-Frame Preservation — SteadyDancer: Harmonized and Coherent Human Image Animation with First-Frame Preservation (published 2025-11-24)
• Time-to-Move: Training-Free Motion Controlled Video Generation via Dual-Clock Denoising — Time-to-Move: Training-Free Motion Controlled Video Generation via Dual-Clock Denoising (published 2025-11-11)
• MotionStream: Real-Time Video Generation with Interactive Motion Controls — MotionStream: Real-Time Video Generation with Interactive Motion Controls (published 2025-11-03)
• LTX Studio Elements — tag, edit, and combine scene elements for visual consistency — LTX Studio Elements — tag, edit, and combine scene elements for visual consistency (published 2025-11-03)

2025-10

• Generative View Stitching — Generative View Stitching (published 2025-10-28)
• Video-As-Prompt: Unified Semantic Control for Video Generation — Video-As-Prompt: Unified Semantic Control for Video Generation (published 2025-10-23)

2025-09

• NewtonGen: Physics-Consistent and Controllable Text-to-Video Generation via Neural Newtonian Dynamics — NewtonGen: Physics-Consistent and Controllable Text-to-Video Generation via Neural Newtonian Dynamics (published 2025-09-25)
• Wan-Animate: Unified Character Animation and Replacement with Holistic Replication — Wan-Animate: Unified Character Animation and Replacement with Holistic Replication (published 2025-09-17)
• SpatialVID: A Large-Scale Video Dataset with Spatial Annotations — SpatialVID: A Large-Scale Video Dataset with Spatial Annotations (published 2025-09-11)

2025-07

• Cameras as Relative Positional Encoding — Cameras as Relative Positional Encoding (published 2025-07-14)
• Moonvalley Marey — controllable AI video model with 360° camera, pose, and motion transfer — Moonvalley Marey — controllable AI video model with 360° camera, pose, and motion transfer (published 2025-07-08)

2025-06

• Hunyuan-GameCraft: High-dynamic Interactive Game Video Generation with Hybrid History Condition — Hunyuan-GameCraft: High-dynamic Interactive Game Video Generation with Hybrid History Condition (published 2025-06-20)
• Sekai: A Video Dataset towards World Exploration — Sekai: A Video Dataset towards World Exploration (published 2025-06-18)
• Video World Models with Long-term Spatial Memory — Video World Models with Long-term Spatial Memory (published 2025-06-05)
• FullDiT2: Efficient In-Context Conditioning for Video Diffusion Transformers — FullDiT2: Efficient In-Context Conditioning for Video Diffusion Transformers (published 2025-06-04)
• UNIC: Unified In-Context Video Editing — UNIC: Unified In-Context Video Editing (published 2025-06-04)
• Voyager: Long-Range and World-Consistent Video Diffusion for Explorable 3D Scene Generation — Voyager: Long-Range and World-Consistent Video Diffusion for Explorable 3D Scene Generation (published 2025-06-04)
• CamCloneMaster: Enabling Reference-based Camera Control for Video Generation — CamCloneMaster: Enabling Reference-based Camera Control for Video Generation (published 2025-06-03)
• Context as Memory: Scene-Consistent Interactive Long Video Generation with Memory Retrieval — Context as Memory: Scene-Consistent Interactive Long Video Generation with Memory Retrieval (published 2025-06-03)

2025-05

• CameraBench: Towards Understanding Camera Motions in Any Video — CameraBench: Towards Understanding Camera Motions in Any Video (published 2025-05-01)

2025-04

• GenDoP: Auto-regressive Camera Trajectory Generation as a Director of Photography — GenDoP: Auto-regressive Camera Trajectory Generation as a Director of Photography (published 2025-04-09)

2023-12

• WonderJourney: Going from Anywhere to Everywhere — WonderJourney: Going from Anywhere to Everywhere (published 2023-12-06)