The independent claim of US12677106B1, a patent granted on July 7, 2026 and assigned to Meta Platforms Technologies, LLC, is directed to a familiar hearing problem: following one voice when several people are talking at once. The record calls its inputs a plurality of sources, and the claimed method processes each of them before mixing them back into a single spatial rendering aimed at the listener's two ears. What follows is a plain-terms walk through what the granted claim actually recites, the classification it carries, and the cohort it issued alongside. To be clear at the outset, this is an issued patent, not a pending application.

Claim 1 begins by receiving audio data from a plurality of sources. Then, for each source, it performs two calculations. The first is a pitch similarity, and the claim specifies that this value is determined in real time to dynamically adjust perceptual audio cues. Those cues are enumerated in the claim itself as one or more of whispered backgrounds, time-dilated vowels, and enhanced sound onsets. The second calculation is an interaural level difference, computed based at least in part on an attenuation level or a dynamic range of interaural time differences. The method then generates spatial audio from the pitch similarity or the interaural level difference, and causes output of audio through one or more speakers based on that generated spatial audio. Here is the claim in full:

A computer-implemented method for generating spatial audio, comprising: receiving audio data from a plurality of sources; for each source of the plurality of sources: calculating a pitch similarity for the audio data, the pitch similarity determined in real time to dynamically adjust perceptual audio cues, wherein the perceptual audio cues comprise one or more of whispered backgrounds, time-dilated vowels, and enhanced sound onsets; and calculating an interaural level difference based at least in part on at least one of an attenuation level or a dynamic range of interaural time differences of the audio data; generating spatial audio based at least in part on at least one of the pitch similarity of the audio data or the interaural level difference of the audio data; and causing output of audio, through one or more speakers, based at least in part on the generated spatial audio.— Statistical provisioning of perceptual audio cues to enhance speech, US12677106B1

Reading the claim in plain terms

Stripped to its structure, the claim describes a per-source pipeline. Every talker or sound the system separates out gets its own pitch analysis and its own interaural measurement before anything is recombined. The pitch-similarity step is the claim's psychoacoustic core: rather than simply making a target voice louder, the method reshapes the perceptual character of the mix, and the claim ties that reshaping to three named manipulations. Whispered backgrounds pull competing sources back; time-dilated vowels stretch the parts of speech that carry intelligibility; enhanced sound onsets sharpen the transients the auditory system uses to lock onto a talker. The supporting disclosure adds concrete texture the claim itself keeps general, including a pitch-similarity band-pass in the 200 Hz to 2500 Hz range and vowel time-dilation on the order of ten percent, with audio captured through the microphones of a user's head-mounted display.

The interaural side of the claim is what makes the output spatial rather than merely filtered. Interaural level difference and interaural time difference are the two cues the human system uses to localize sound between the left and right ears, and the claim computes the level difference from an attenuation level or from the dynamic range of the time differences. Because the sources in the disclosed setting may be different entities in an artificial-reality environment, the recombined spatial audio can place each processed voice where the scene puts its speaker, so a listener follows a target talker both by its reshaped timbre and by where it sits in space. Notably, the claim does not require both calculations to drive the output on their own: the generating step recites the pitch similarity or the interaural level difference, so either cue, or both, can inform the spatial audio that is ultimately produced.

Classification and cohort

The record carries CPC classification H04S 7/303, the subclass for stereophonic systems with processing of the spatial sound field, alongside G10L 21/0232 (speech enhancement by frequency-domain processing), G10L 25/90 (pitch determination), H04R 3/04, and H04S 3/008. That combination places the grant squarely at the intersection of spatial-audio rendering and speech-enhancement signal processing rather than in either alone, which matches a claim that treats localization cues and speech intelligibility as one coupled problem.

US12677106B1 issued within a same-week Reality Labs cohort of granted patents. On the audio side, US12669974B1 is directed to creating custom audio mixes for artificial-reality environments, and US12676139B1 covers mixed-reality text narration that reads on-screen text aloud and re-syncs when the text changes. The hardware around the ears and eyes shows up too: US12674991B2 claims augmented-reality glasses temple-arm components with a speaker sited between front and rear battery cells, while US12663619B2 is directed to a wide field-of-view optical lens assembly with a piezo-actuated tunable lens. Interaction gets its own grant in US12663864B2, which claims EMG-based wrist control for interacting with vehicles.

Taken together as a matter of coverage, the granted claims describe a Reality Labs stack in which audio, optics, and wearable interaction each carry issued patents, and the speech-enhancement grant is the audio-perception member of that set. The claim reported here is factual coverage as issued: a per-source method coupling real-time pitch analysis to named perceptual manipulations and interaural cues, granted July 7, 2026, and classified under the spatial-audio and speech-enhancement CPC subclasses noted above.