Open ear earbuds have always shared one inherent structural weakness. Because they never seal the ear, their microphones have no acoustic shelter: the wearer’s voice and the noise around them reach the array through the same medium, air, arriving as one inseparable signal. Filtering has been the only available answer, and filtering degrades as the environment grows louder. Today, Anker Innovations announced Anker soundcore AeroClip 2, which no longer relies on air alone. Two voice pickup units capture the mechanical vibrations of the wearer’s own vocal cords, a signal physically coupled to the voice and structurally independent of airborne noise, and the THUS AI chip cross references that stream against the signal from four MEMS microphones in real time. Anker calls the result “Loud World, Clear You.”
The Microphone Problem in Open-Ear Audio
Conventional open ear earbuds separate speech from surroundings using microphone arrays alone. Because both signals share one medium, the isolation problem is not better microphones but insufficient information: no array can recover what the air has already mixed together. Filtering aggressively is the only compensation, which is why open ear call quality has historically degraded in traffic, transit, and open plan offices.
AeroClip 2 introduces a second, independent source of information. Its 6 sensor hybrid array pairs four high signal to noise ratio (SNR) MEMS microphones with two voice pickup units that register vibration transmitted through the wearer’s body rather than through the air around them. Because that vibration originates at the vocal cords, it carries the voice and not the room.
Reading both streams at once, in real time, is not a filtering task but an inference task, and it has to run on the earbud itself, within a power budget a wearable can spare. AeroClip 2 executes it on THUS, which delivers 4.6 GFLOPS, approximately 150 times the AI computing power of the previous generation.
THUS™: A Chip Platform Already Proven in the Ear
The chip AeroClip 2 runs on is not new, and that is the point: THUS has already been tuned, shipped, and independently measured in an earbud before arriving here.
Every chip built today descends from an architecture established in 1945 by mathematician John von Neumann: break a problem into discrete steps, and let memory and processor sit apart to handle them one at a time. AI breaks that assumption. A neural network processes end to end, drawing on millions of parameters at once, and every inference sends all of them on a round trip between memory and processor. In a data centre, that movement is a cost. In an earbud powered by a battery smaller than a fingernail, it is a wall: over 90 percent of chip power spent moving data before a single calculation begins.
Nature had already solved it: neurons store and process information in the same place. THUS applies that principle to silicon, embedding computation directly inside NOR Flash memory cells so parameters never have to move, freeing that energy for computation instead. NOR Flash compute in memory also occupies roughly one sixth the footprint of SRAM alternatives, small enough for the most space constrained wearables.
That platform has already been validated in a shipping product. Anker soundcore Liberty 5 Pro, the first earbud tuned to run THUS, had the result measured externally rather than claimed internally: in April 2026 it earned Guinness World Records certification for the highest objectively measured speech quality score in TWS earbud history. AeroClip 2 brings that same proven silicon to an open ear design, earning World’s Clearest Open Ear Clip On Earbuds for Calls in Sep 2026, where the acoustic problem is harder and the chip has more to do.
Press Release by Anker Singapore