Sony · AI / ML · Filed Oct 9, 2025 · USPTO published Sep 24, 2026 · Sep 24 release

Sony Patents an AI That Scores How Well Creative Building Blocks Fit Together

When you're building something creative, knowing which pieces belong together is half the battle. Sony is patenting an AI model trained to score exactly that: how well one element of a piece of content relates to another.

By Patentlyze Team · Updated Sep 24, 2026 · 3 min read

An animation work is broken down into its constituent parts, including a script with cast members and two-dimensional animation characters. Drawing from patent filing US 2026/0289403 A1. — An animation work is broken down into its constituent parts, including a script with cast members and two-dimensional animation characters.All 23 drawings ↓

Publication number: US 2026/0289403 A1(US20260289403A1)
Applicant: SONY GROUP CORPORATION
Filed · Published: Oct 9, 2025 · Sep 24, 2026
Inventors: SUSUMU TAKATSUKA, ITARU SHIMIZU, HIROKI TETSUKAWA
Status when we published: Waiting for an examiner
Document: 20 claims
US classification: 706/12
Parent application: National Stage Entry of PCT/JP2024/011697 (filed Mar 25, 2024)(PCTJP2024011697)

AI training & infrastructure

What Sony's content-relevance AI actually does for creators

Ever tried to decide whether a background score fits the mood of a scene, or whether a lyric matches a melody? That judgment call is something Sony wants to hand off to an AI.

The patent describes a system that breaks creative works (think songs, videos, or other media) into their individual parts, then trains a model to score how relevant any one part is to another. The output is a number, an index value, that tells you how well component A pairs with component B.

The idea is to give creators a tool that can suggest, check, or rank combinations rather than leaving every pairing decision entirely to instinct. Sony frames this as a creation-support system, something sitting alongside a human artist rather than replacing them.

From the filing · CLAIM 1

acquiring a plurality of components constituting each of a plurality of different pieces of content; and generating a learning model that outputs an index value indicating relevance of a first component with respect to a second component among the plurality of components.

Translation: The system studies different creative works to figure out how well their building blocks fit together.

How the model learns relevance across content components

The system starts by ingesting a large set of existing content pieces and breaking each one down into its components (individual elements like a musical phrase, a visual clip, a text line, or any other building block depending on the media type).

From that dataset, a learning model is trained. Its job is to output an index value (a numeric score) indicating how relevant a first component is with respect to a second component. Higher relevance presumably means the two elements belong together or work well in combination; lower relevance means they clash or are unrelated.

The patent also references a presentation method and an evaluation method, suggesting the system doesn't just generate scores internally but can surface those scores to a user and apply them to judge or rank candidate pairings. A generation method is also mentioned, hinting that the scored relevance can feed into actually producing new content suggestions.

The claim language is intentionally broad: "components" and "content" are left undefined, which means the approach could apply to music, video, text, games, or any structured creative medium Sony chooses.

From the filing · THE ABSTRACT

The technique according to the present disclosure can be applied to a system related to creation support for a creator, for example.

Translation: Sony built this AI tool to help artists brainstorm and assemble their creative projects.

What this means for AI-assisted creative tools

For anyone using a Sony creative tool, this could mean getting an automated second opinion on whether the pieces of a project fit together before you commit. That kind of feedback is normally something you get from a collaborator or a trained ear; an AI that approximates it could save real time during early drafts.

The broader relevance is that Sony keeps filing on AI-assisted creation tools, and a relevance-scoring model is a foundational layer. If it works, it could sit underneath recommendation engines, auto-arrangement tools, or content search features in Sony's music, film, or game production software.

Sony's 13th filing in the AI training and infrastructure work we've tracked since June adds to a run that includes a game-based matchmaking patent and one on label-free network design.

Editorial take

From a reader-impact standpoint, the core question is: would you actually notice this in a product? A relevance score sounds useful in the abstract, but the patent gives almost no detail about what counts as a component, how the training data is assembled, or how accurate the scoring needs to be to beat a creator's own instinct.

The claim is about as wide as a patent claim can get. That breadth might be a legal strategy, but it also means there's very little to evaluate technically. The invention, as written, is closer to a framework than a finished tool.

Sony is clearly interested in the creation-support space, and a model that scores element compatibility could be a real building block for music or video software. But this filing, on its own, is early-stage infrastructure thinking, not something you'd expect to find in a product update anytime soon.

Source: US 2026/0289403 A1 at the USPTO (PDF). Full patent text and figures are in the official publication.

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.

Add Patentlyze on Google One click.

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.