Frame Interpolation
Frame interpolation is the use of AI models to generate intermediate frames between existing ones, used in animation to smooth motion, change frame rate, or recover slow-motion from standard-rate footage.
In a traditional pipeline, this work is done by a human in-betweener: the keyframes are drawn, the in-between drawings smooth the motion. Modern AI tools can generate convincing in-betweens automatically, with quality good enough for production in many use cases.
Inside our pipeline, frame interpolation is used in two places. First, to up-rate AI-generated video from the model's native output (often 24 fps) to broadcast frame rates (50 or 60 fps for slow-motion shots). Second, in 2D animation, to fill in-betweens between hand-drawn keys, leaving the hand-keyed work for the moments that matter most.
Limits: the model can guess wrong on fast motion or occlusion, producing visible artefacts. Production work pairs interpolation with a clean-up pass and, for hero shots, a hand-drawn check. On Inchstones, the hero in-betweening stays hand-keyed; on faster turnaround work, AI in-betweens are part of the stack.
Frame interpolation also matters for delivery specs. Different broadcasters require different frame rates; AI-driven conversion is often cleaner than older optical-flow tools.
Related
Related services
Sources
Academic papers, recognised industry standards, and canonical industry texts that back up claims in this entry.
- Enhanced Deep Animation Video Interpolation. Liu, Xu, Wang et al., arXiv, 2022Supports: AI animation frame interpolation
- Neural Frame Interpolation for Rendered Content. Briedis, K. M., Djelouah, A., Meyer, M., McGonigal, I., Gross, M., Schroers, C., Disney Research / ACM SIGGRAPH Asia, 2021Supports: Production-grade neural frame interpolation for rendered content
- Real-Time Intermediate Flow Estimation for Video Frame Interpolation. Huang, Z., Zhang, T., Heng, W., Shi, B., Zhou, S., arXiv (ECCV), 2022Supports: Optical-flow-based frame interpolation method (RIFE)
Frequently asked questions
Does AI interpolation replace in-betweeners?
On hero shots, no. The shape of the motion, the personality of the in-between, and the timing nuance are still hand-decided. On faster work or large volumes, AI takes the routine in-betweens and animators direct the keys. The job changes shape, the craft remains.
Is this the same as motion smoothing on a TV?
Same idea, different bar. Consumer motion smoothing on TVs uses cheap, real-time interpolation. Production AI interpolation runs slower and produces results clean enough for delivery. The visual result is in a different league.
What about frame rate conversion?
AI-driven conversion is now the default in most pipelines for 24 to 50 or 60 fps conversion. It handles the panning and motion artefacts more cleanly than older tools. For broadcast deliverables where frame rate must change between regions, this is the workflow.