Sony Files Patent to Grade How Naturally an Animated Character Switches Moves
When an animated character jumps from one recorded move to another, the join often looks broken: a foot teleports, or the whole body lurches. Sony's filing describes software that generates the in-between movement and then grades how natural it looks.
What Sony's animation transition scoring does
A video game character is strolling across a stage when the player tells it to kick a ball. The walking and the kicking were recorded separately, so if software simply snaps from one to the other, a leg can seem to teleport and the whole scene looks broken.
Sony's patent application describes a tool that fills that gap and then grades its own work. An AI creates the in-between movement, and the software checks for the flaws your eye would catch right away: a sudden burst of speed, a jerky lurch, or a foot that skates across the floor instead of planting.
Instead of handing you a single result, the tool makes many versions and lines them up from best score to worst. You, or the animator using it, pick the one that looks right, and the software stitches it between the two original clips.
circuitry configured to obtain interpolation motion data that interpolates between first motion data and second motion data that are temporally and spatially independent …
Translation: The system blends two completely different character movements together smoothly.
How Sony grades each candidate transition
The filing starts with two pieces of motion data, which are digital skeleton recordings that track where a character's joints are over time. They come from motion capture or from hand animation. A generative model (an AI trained to produce new content) creates the missing in-between movement. The filing says it is trained by cutting existing recordings into three chunks, hiding the middle one, and teaching the model to recreate it.
The key step is the scoring. The software calculates several checks, each tied to how fast the body moves:
- Speed match: the fastest moment of the in-between motion and its average speed are compared with the speed of the frames before and after it.
- Jerk: how abruptly the acceleration changes, since sudden speeding up or braking looks unnatural.
- Foot sliding: whether a toe that touches the ground keeps moving along it.
- Use-case fit: whether the generated motion suits the job, for example not filling a gap in one performer's motion with a very different body type's movement.
Each check is multiplied by a weighting factor and added into one total score.
The tool then changes settings such as the time gap, start and end postures, number of steps, style labels like "dance-like," and a random seed. Some are set automatically from the root joint (the waist) speed and position. It repeats until it has a preset number of candidates, sorts them by score, and lets the user choose.
The filing also describes using this to repair gaps in motion capture caused by blocked sensors.
… calculate an index related to motion velocity of each motion data based on the first motion data, the second motion data, and the obtained interpolation motion data …
Translation: It measures how fast the character moves before, during, and after the transition.
Why this matters for animators and avatars
For animators and avatar creators, the practical gain is a shorter path to a convincing transition. Generated motion comes with many settings, and the filing says it is hard for a person to set them all by hand. A ranked shortlist lets someone judge a handful of options instead of tuning each setting. The examiner's first review rejected the claims as filed, which is a routine stage that is usually followed by amendments.
The filing also points to repairing motion capture recordings where a sensor lost track of a performer, and to blending live motion with pre-recorded gestures for avatars in VR games and virtual spaces. It names no product, so treat this as a description of a technique and not a preview of a release.
Sony's 636th filing in our Sony coverage since May follows work on cutting cloud game lag and eye-tracking AI training.
You will probably never see this tool. You would see its result: a character that moves from a walk to a kick without a visible glitch, which is exactly the flaw people notice in games, avatar streams and animated videos.
The useful idea is the scoring. Checking for sudden speed jumps, jerky lurches and sliding feet gives software a way to judge smoothness roughly the way your eye does. That spares a creator from watching dozens of options one by one.
The weak spot is taste. The filing itself says what looks right depends on the viewer's preference, so a ranked list still needs a human to make the final call. Overall, the tool is a sensible, modest improvement to an animator's workflow, and most players would only notice it as the absence of awkward moments.
Get our take in your Top Stories
Liked this breakdown? Add Patentlyze as a preferred source on Google, and our plain-English take shows up more often in your Top Stories the next time Sony patent news breaks.
There are more where this came from
We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.
The drawings
8 drawing sheets from US 2026/0330735 A1 · click any drawing to enlarge
Want this weekly breakdown for a company we don't cover? Patentlyze Pro →
Be the first to weigh in