Frame Interpolation / Inbetweening¶
Animation & Motion
Open Source¶
-
Bridging Event Streams and DiT: Event-Guided Video Frame Interpolation Discovered by AI
2026
This paper introduces an event-guided video frame interpolation framework that leverages high-temporal-resolution motion cues from event cameras to guide latent diffusion models, resolving motion blur and structural distortions without training an event-assisted model.
diffusion-modelevent-cameravideo-interpolation -
Following Motion for Sequential Modeling in Video Frame Interpolation Discovered by AI
2026
Motion-Guided Mamba for Video Frame Interpolation (MGMVFI) adapts the selective state space model (Mamba) to video frame interpolation by incorporating dynamic motion trajectories to overcome the limitations of predefined scanning orders.
mambastate-space-modelvideo-interpolation -
MASS: Motion-Aligned Selective Scan for Refinement in Flow-Based Video Frame Interpolation Discovered by AI
2026
Motion-Aligned Selective Scan (MASS) is a video frame interpolation framework that reformulates feature scanning in selective State Space Models (SSMs) to align with physical motion, improving interpolation under large, non-linear motions and complex occlusions.
mambastate-space-modelvideo-interpolation -
SNM-VFI: Symmetric Nonlinear Motion-Guided Generative Video Frame Interpolation Discovered by AI
2026
SNM-VFI is a training-free framework for motion-controllable generative video frame interpolation that guides the generative process of pre-trained video diffusion models using correspondence-aware frames from a symmetric nonlinear motion model.
diffusion-modeloptical-flowvideo-interpolation -
SPEED: One-Step Pixel Diffusion for High-quality Video Frame Interpolation Discovered by AI
2026
A one-step pixel diffusion framework for high-quality video frame interpolation. It uses a progressive multi-stage architecture with dynamic patch scaling to learn multi-scale motion, structural, and appearance representations directly in pixel space, avoiding the detail loss of latent diffusion and the high latency of multi-step sampling.
diffusioninterpolationone-steppixel-space -
EDEN Discovered by AI
2025
A state-of-the-art diffusion-based video frame interpolation model designed to handle complex and large-motion scenarios. It utilizes a transformer-based token design to generate sharp, temporally consistent intermediate frames without the typical blurriness of flow-based methods.
diffusionframe-interpolationlarge-motion -
Hierarchical Flow Diffusion for Efficient Frame Interpolation Discovered by AI
2025
A video frame interpolation method that models bilateral optical flow explicitly using hierarchical diffusion models. By reducing the search space in the denoising procedure and using a flow-guided image synthesizer, it achieves state-of-the-art accuracy while being over 10 times faster than previous diffusion-based methods.
diffusionefficiencyoptical-flowvideo-interpolation -
MoG-VFI Discovered by AI
2025
A generative video frame interpolation framework that bridges the gap between flow-based motion stability and generative flexibility by integrating explicit motion guidance.
animationdiffusioninbetweeningmotion-aware -
MoMo Discovered by AI
2025
A diffusion-based video frame interpolation approach that disentangles motion modeling from pixel-space generation, achieving high perceptual quality with reduced computational costs.
diffusioninbetweeninginterpolationvfi -
PoseFuse3D-KI: Controllable Human-centric Keyframe Interpolation with Generative Prior Discovered by AI
2025
A framework that integrates 3D human guidance signals (such as SMPL-X) into a video diffusion model for controllable human-centric keyframe interpolation, resolving issues with complex, articulated human motions.
diffusioninterpolationkeyframesmpl-x -
SAGE: Structure-Aware Generative Video Transitions between Diverse Clips Discovered by AI
2025
A structure-aware generative video transition framework designed to synthesize intermediate frames between diverse video clips with large temporal gaps or significant semantic differences.
generativeinterpolationtransition -
SG-RIFE Discovered by AI
2025
A real-time video frame interpolation framework that bridges the gap between flow-based methods and diffusion-based models by injecting dense semantic priors from DINOv3. It achieves diffusion-competitive perceptual quality while maintaining near real-time speeds.
diffusioninterpolationreal-timerife -
Time-adaptive Video Frame Interpolation based on Residual Diffusion Discovered by AI
2025
A diffusion-based video frame interpolation method designed specifically for traditional hand-drawn animation. It explicitly handles and re-estimates interpolation time to cope with large temporal variations and non-linear motion.
animationanimediffusioninterpolation -
TLB-VFI: Temporal-Aware Latent Brownian Bridge Diffusion for Video Frame Interpolation Discovered by AI
2025
A temporal-aware latent Brownian bridge diffusion model (TLB-VFI) for video frame interpolation, designed to be highly efficient with fewer parameters and less training data while maintaining high temporal coherence.
diffusioninterpolationtemporal-coherence -
ToonComposer: Streamlining Cartoon Production with Generative Post-Keyframing Discovered by AI
2025
ToonComposer is a generative post-keyframing model that unifies 2D cartoon inbetweening and colorization, synthesizing smooth, temporally coherent frame sequences directly from sparse keyframes under large non-rigid motions.
cartooncel-animationinbetweeningkeyframe-interpolation -
BiM-VFI: Bidirectional Motion Field-Guided Frame Interpolation for Video with Non-uniform Motions Discovered by AI
2024
A frame interpolation model designed for videos with non-uniform motions (such as acceleration, deceleration, and direction changes) that resolves time-to-location ambiguity and reduces blur using bidirectional motion fields.
frame-interpolationmotion-estimationvfi -
Frame Interpolation with Consecutive Brownian Bridge Diffusion Model Discovered by AI
2024
A diffusion-based video frame interpolation model that proposes a consecutive Brownian Bridge diffusion process to reduce the cumulative variance of generated latent representations, achieving high-fidelity results.
brownian-bridgediffusion-modelframe-interpolation -
Framer Discovered by AI
2024
An interactive frame interpolation framework that allows users to customize the transition process between two keyframes by tailoring the trajectories of selected keypoints.
animationinbetweeninginteractivekeypoint -
Generalizable Implicit Motion Modeling for Video Frame Interpolation Discovered by AI
2024
Introduces Generalizable Implicit Motion Modeling (GIMM), which encodes spatiotemporal motion latents from bidirectional flows to effectively represent input-specific motion priors for frame interpolation.
interpolationmotion-modelingvfi -
Generative Inbetweening through Frame-wise Conditions-Driven Video Generation Discovered by AI
2024
A generative inbetweening framework that utilizes frame-wise conditions to guide video diffusion models, significantly improving temporal stability and motion coherence between two input keyframes.
diffusioninbetweeningkeyframe -
Generative Inbetweening: Adapting Image-to-Video Models for Keyframe Interpolation Discovered by AI
2024
Adapts a pre-trained image-to-video diffusion model for keyframe interpolation by predicting and fusing forward and backward video sequences conditioned on start and end frames.
diffusioninbetweeningkeyframe -
InterpAny-Clearer Discovered by AI
2024
A video frame interpolation framework that addresses the velocity ambiguity problem in non-uniform motions (such as acceleration and deceleration) by introducing a new indexing method, resulting in significantly clearer interpolated frames.
eccvinbetweeninginterpolationvfi -
Joint Stroke Tracing and Correspondence for 2D Animation Discovered by AI
2024
A joint stroke tracing and correspondence framework for 2D animation inbetweening. It takes consecutive raster keyframes and a single vector drawing from the starting keyframe as inputs, generating vector images for the remaining keyframes with one-to-one stroke correspondence compatible with standard 2D animation workflows.
2d-animationinbetweeningstroke-correspondencevector-animation -
LayerInbetween: Occlusion-Aware Stroke Correspondence and Inbetweening with Automatic Layering Discovered by AI
2024
An occlusion-aware framework for vector stroke correspondence and automatic inbetweening in 2D animation. It performs automatic layering to guide stroke tracing and correspondence finding for occluded strokes, resolving occlusion in the inbetween frames.
animationinbetweeningstrokevector -
LDMVFI Discovered by AI
2024
A generative video frame interpolation method based on latent diffusion models that approaches the interpolation task from a perceptual quality perspective rather than relying solely on pixel-level reconstruction losses.
diffusioninbetweeninginterpolationvfi -
PerVFI Discovered by AI
2024
A perception-oriented video frame interpolation method that addresses blur and ghosting artifacts caused by motion errors. It utilizes asymmetric blending and a normalizing flow-based generation module to synthesize sharp, realistic intermediate frames.
interpolationperception-orientedvideo-interpolation -
SGM-VFI Discovered by AI
2024
A video frame interpolation pipeline designed to handle large motion by integrating global-level matching information to alleviate the limitations of local receptive fields.
inbetweeninginterpolationlarge-motionvfi -
SPEED: One-Step Pixel Diffusion for High-quality Video Frame Interpolation Discovered by AI
2024
SPEED is a one-step pixel diffusion framework for high-quality video frame interpolation. It utilizes a progressive multi-stage architecture with dynamic patch scaling to effectively capture multi-scale motion, structural, and appearance representations in a single step.
diffusioninterpolationone-stepvfi -
Thin-Plate Spline-based Interpolation for Animation Line Inbetweening Discovered by AI
2024
A specialized line art inbetweening method that utilizes thin-plate spline-based transformation to model keypoint correspondences and handle large motions and line disconnections in animation line drawings.
animationinbetweeninglinearttps -
ToonCrafter: Generative Cartoon Interpolation
2024
Diffusion-based generative interpolation that synthesizes in-between frames for cartoon/anime keyframes, handling large non-linear motion and occlusion.
animediffusioninterpolation -
VFIMamba: Video Frame Interpolation with State Space Models Discovered by AI
2024
The first approach to adapt State Space Models (SSM/Mamba) to video frame interpolation, utilizing a Mixed-SSM Block for efficient inter-frame modeling and high-resolution synthesis.
interpolationmambavfi -
ViBiDSampler: Enhancing Video Interpolation Using Bidirectional Diffusion Sampler Discovered by AI
2024
A bidirectional diffusion sampling strategy for keyframe interpolation that adapts single-frame image-to-video models to two-frame (start and end) conditioned generation, using CFG++ and Dual-directional Diffusion Sampling (DDS) to ensure temporal coherence.
bidirectionaldiffusioninterpolationkeyframe -
AMT: All-Pairs Multi-Field Transforms for Efficient Frame Interpolation Discovered by AI
2023
A lightweight, fast, and accurate video frame interpolation model that utilizes bidirectional correlation volumes for all pairs of pixels and multi-field transforms to retrieve correlations and refine flows.
frame-interpolationoptical-flowvfi -
AnimeInbet Discovered by AI
2023
A deep learning framework designed specifically for the inbetweening of black-and-white cartoon line drawings. It uses vertex geometric embedding, a vertex correspondence Transformer, and a visibility predictor to synthesize clean and complete intermediate line drawings, outperforming raster-based methods on large motions.
animeinbetweeninginterpolationlineart -
Bridging the Gap: Sketch-Aware Interpolation Network for High-Quality Animation Sketch Inbetweening Discovered by AI
2023
SAIN is a sketch-aware frame interpolation network designed for 2D animation hand-drawn sketch inbetweening, effectively preserving thin structural lines and handling large non-rigid motion changes between keyframes.
2d-animationframe-interpolationline-artsketch-inbetweening -
ComfyUI-Frame-Interpolation Discovered by AI
2023
A comprehensive suite of custom nodes for ComfyUI that integrates various state-of-the-art video frame interpolation models, including RIFE, FILM, and AMT, into generative workflows.
comfyuiinterpolationvfi -
Extracting Motion and Appearance via Inter-Frame Attention for Efficient Video Frame Interpolation Discovered by AI
2023
An efficient video frame interpolation model that leverages inter-frame attention to extract motion and appearance information in a unified, lightweight hybrid CNN-Transformer architecture.
attentioninterpolationvfi -
GMFSS Discovered by AI
2023
An open-source video frame interpolation framework based on GMFlow, specifically optimized for anime and cartoon footage to produce smooth motion transitions.
animeframe-interpolationgmflowvfi -
GMFSS Fortuna Discovered by AI
2023
A highly popular, community-developed video frame interpolation model based on GMFlow, specifically optimized and fine-tuned for 2D anime and animation production.
animegmflowvfi -
FILM Discovered by AI
2022
A unified single-network frame interpolation model designed to synthesize multiple intermediate frames from two input images with large in-between motion. It features a multi-scale feature extractor that shares weights across scales and is trainable from frame supervision alone.
interpolationlarge-motionvideo-interpolation -
IFRNet Discovered by AI
2022
An efficient encoder-decoder based network designed for fast and real-time intermediate frame synthesis. It refines bilateral intermediate flow fields and bilateral intermediate features jointly, achieving a great balance between interpolation quality and inference speed.
efficientframe-interpolationreal-time -
RIFE: Real-Time Intermediate Flow Estimation
2022
A fast flow-based frame interpolation model, widely used (with anime-tuned variants) to generate in-between frames and smooth low-frame-rate animation.
interpolationreal-time -
AnimeInterp Discovered by AI
2021
An animation-specific video frame interpolation framework that addresses challenges like non-linear motion and large displacements. It utilizes segment-guided matching, recurrent flow refinement, and warping-synthesis modules to produce high-quality intermediate frames for cartoon videos.
animeinterpolationvideo-interpolation -
AnimeInterp: Deep Animation Video Interpolation
2021
Segment-guided frame interpolation designed for the flat colors and large motions of cartoon/anime, addressing failure modes of photoreal interpolators.
animeinterpolation -
EISAI Discovered by AI
2021
A 2D animation frame interpolation system focused on improving perceptual quality. It introduces SoftsplatLite (SSL) for forward-warping interpolation and a perceptual-loss-driven training scheme to handle extreme motion and preserve sharp hand-drawn lines.
animeinterpolationperceptual-quality -
Flowframes Discovered by AI
2021
A popular open-source Windows GUI application for video frame interpolation, supporting advanced AI models like RIFE, DAIN, and FLAVR to convert low-framerate videos into smooth high-framerate footage.
dainguiinterpolationrife -
FLAVR: Flow-Agnostic Video Representations for Fast Frame Interpolation Discovered by AI
2020
FLAVR is a flow-agnostic video frame interpolation architecture that uses 3D space-time convolutions to directly predict intermediate frames without relying on optical flow estimation or warping.
3d-cnnframe-interpolationslow-motionvideo -
DAIN Discovered by AI
2019
A classic depth-aware video frame interpolation model that explicitly detects occlusions by exploring depth information. It introduces a depth-aware flow projection layer to synthesize intermediate flows and frames, making it highly effective for complex scenes with overlapping objects.
depth-awareframe-interpolationoptical-flow
Proprietary / Industry Tools¶
-
CACANi Discovered by AI
2022
A professional 2D vector animation software that specializes in computer-assisted inbetweening and coloring. It automatically generates intermediate frames from keyframes while maintaining stroke correspondence, significantly reducing the manual workload for clean-up and inbetween artists.
animeinbetweeningsoftwarevector -
Squirrel Video Frame Interpolation (SVFI) Discovered by AI
2021
A popular consumer-grade AI video frame interpolation and enhancement application widely used by anime creators and video editors to convert low-FPS footage into high-framerate videos using models like RIFE and GMFSS.
animationappinbetweeningrife -
SVFI Discovered by AI
2021
A popular consumer-grade video frame interpolation and upscaling software widely used by anime fans and video creators. It integrates various AI models (such as RIFE and GMFSS) to convert low-framerate videos and image sequences into smooth, high-framerate footage.
anime-upscalingguisoftware -
SVFI (Squirrel Video Frame Interpolation) Discovered by AI
2021
A popular commercial video frame interpolation and super-resolution software tool widely used by video editors and anime creators. It integrates advanced AI models like RIFE and GMFSS to upscale and smooth out low-FPS animations.
animeappinbetweeninginterpolation -
SmoothVideo Project Discovered by AI
2011
A well-established real-time frame interpolation software that increases video frame rates during playback. It utilizes GPU acceleration and advanced motion-vector algorithms to interpolate standard videos to 60 FPS, 120 FPS, or higher in real-time.
playbackreal-timesoftware
