Home Robot Noise Specs: What Quiet Really Means

A robot vacuum can be scheduled for noon. A lawn robot can run outside. A home humanoid is different. If it is meant to fetch objects, tidy rooms, respond to voice, patrol at night, or assist an older adult, it may be moving near people for hours. That makes sound less like a comfort detail and more like a home-readiness filter.

The reason this topic matters now is simple: 1X NEO is one of the first consumer-facing humanoids to make quiet operation part of the product story. 1X says NEO operates at a noise level quieter than a modern refrigerator. Heise, covering 1X's Hayward production ramp, reported an operating-noise claim of about 22 dB, explicitly noting that it was not dB(A).

That sounds impressive. It also raises the right buyer question: what would a useful, honest home-robot noise spec actually include?

How quiet does a home robot need to be?

There is no single magic number. The right threshold depends on the task, room, time of day, surface, and distance from people. A robot moving a cup at 11 p.m. should be judged differently from a robot vacuum running a turbo cleaning pass at 2 p.m.

A practical buyer framework looks like this:

That last row is easy to miss. A robot can advertise a low steady-state number and still be annoying if it makes sharp motor corrections, drags furniture, slams into thresholds, announces errors loudly, or spins fans up during heavy compute. For homes, the sound profile matters more than a single best-case number.

This is also why dB wording matters. The decibel scale is logarithmic, so small numerical changes can feel larger than they look. dB(A) is weighted to approximate human hearing sensitivity; a plain dB claim may not be comparable. Distance, height, room acoustics, floor type, payload, speed, and microphone placement all change the result. NIOSH uses 85 dBA averaged over an eight-hour workday as an occupational exposure limit, but home comfort is a much stricter problem than hearing safety.

What makes the 1X NEO quiet claim unusual?

Most humanoid marketing leads with hands, gait, AI models, or factory videos. 1X's quiet claim is unusual because it treats the home as an acoustic space, not just a demo stage.

In the ui44 database, NEO is listed as a $20,000 home-focused humanoid in pre-order status, with a 167 cm body, 30 kg weight, roughly 4 hours of battery life, RGB/depth sensing, tactile skin, microphone array, Wi-Fi, Bluetooth, and household-chore ambitions. The official NEO page emphasizes a soft body, tendon-driven actuators, gentle interaction, and quiet operation. Heise connects the reported 22 dB figure to 1X's in-house Tendo Drives, the company's tendon-driven actuator design.

That combination matters. Noise is not isolated from safety. A light robot with soft coverings, compliant actuation, and lower mechanical noise has a better shot at feeling normal around people than a heavier machine that sounds industrial whenever it bends down.

But buyers should not overread one claim. Before treating NEO as a proven quiet home robot, the missing details are important:

• Was the number measured while standing, walking, reaching, carrying, or doing a chore?
• At what distance and height was the microphone placed?
• Was it measured in dB, dB(A), or another weighting?
• Was the robot on carpet, hard floor, tile, or an acoustic test surface?
• Did the test include fan noise, speech output, remote-assistance sessions, and error recovery?
• Is there a separate maximum noise figure for heavy manipulation or fast motion?

A quiet claim is a good sign. A published test protocol would be much better.

The bigger problem: most humanoids do not publish noise at all

The more revealing comparison is not NEO versus a refrigerator. It is NEO versus every other humanoid that wants to be taken seriously around people.

Unitree G1 is one of the most concrete humanoids ordinary buyers can actually evaluate. ui44 lists it as Available, starting at $13,500, with a 132 cm body, about 35 kg weight, roughly 2 hours of battery life, 23 degrees of freedom on the base model, optional EDU configurations, Wi-Fi 6, Bluetooth 5.2, a depth camera, 3D LiDAR, and a four-mic array. Unitree's official page also includes a blunt caution that users should keep a sufficient safe distance from the robot and understand humanoid limits before purchase.

What the public spec set does not give buyers is a comparable home-noise number. That does not make G1 a bad robot. It does mean a buyer cannot confidently answer basic lifestyle questions: can it move during a video call, in an apartment at night, near a nervous dog, or around someone who is sensitive to mechanical noise?

The same pattern appears across better-known humanoid names. Figure 03 is active but not consumer-priced; ui44 tracks it at 173 cm, 61 kg, about 5 hours of battery life, a 20 kg payload, and factory/residential service infrastructure signals from Figure's production updates. Tesla Optimus Gen 2 remains in development with a stated target around $30,000, a 173 cm, 57 kg body, and no official consumer availability. Apptronik Apollo and Agility Digit are enterprise-first humanoids with service, warehouse, manufacturing, and logistics positioning.

For all of them, noise is still either absent, hard to compare, or not central to the public spec sheet.

That gap is acceptable for a factory pilot. It is not acceptable for a robot that claims it will share a home.

Existing home robots show what disclosure should look like

Cleaning and lawn robots are not perfect analogies for humanoids, but they show something useful: when noise affects real owners, mature categories eventually publish numbers or modes.

ui44 tracks Dyson Spot+Scrub Ai with 57.2 dB max quiet-operation language alongside its 18,000 Pa suction, AI stain detection, wet roller mop, and $1,199.99 price. Husqvarna Automower 450X NERA is tracked at 58 dB while handling large lawns, radar obstacle avoidance, optional EPOS wire-free navigation, and €4,999 reference pricing. Sunseeker S4 publishes ≤60 dB(A) on its official page while using LiDAR plus AI vision for wire-free mapping. Mammotion LUBA 2 AWD 5000 is described in ui44 as sub-60 dB while using RTK plus AI vision and all-wheel drive mowing.

Those are mostly appliance-like numbers. They do not mean those robots are silent. They mean buyers can set expectations. A 58 dB mower outside is different from a 58 dB humanoid walking through a nursery. A robot vacuum can be loud while cleaning because it is doing a loud task: suction, brushrolls, water pumps, mop scrubbing, and dock emptying. A humanoid fetching a glass should not sound like a vacuum.

The more subtle example is Narwal Flow 2. ui44 tracks it with a baby mode that can switch to ultra-quiet operation near cribs, plus toy recognition and pet-oriented behavior. Even without a published full sound table, that mode points in the right direction: quietness should be contextual. A robot should know when noise matters.

What a real home-humanoid noise spec should include

A useful home-humanoid spec sheet should not stop at one headline number. It should separate the robot's sound into operating states.

At minimum, buyers should ask for:

1. Idle / standby noise. Does the robot have fan or coil noise while it is simply awake?
2. Walking noise. How loud is slow indoor movement on hard floor, carpet, and thresholds?
3. Arm movement noise. What happens while reaching into a shelf, lifting a cup, or closing a cabinet?
4. Carrying-load noise. Does actuator noise increase when the robot carries 1 kg, 5 kg, or a near-limit payload?
5. Fan and compute noise. Does vision-language-action inference, mapping, or remote-assistance video processing spin up audible cooling?
6. Speech and alert volume. Can the robot whisper, use haptics/app alerts, or obey quiet hours?
7. Error and recovery noise. Are failed grasps, stuck motions, and docking attempts sharp, repetitive, or manually adjustable?
8. Maximum event noise. What is the loudest normal sound: a footstep, servo correction, gripper slip, fall arrest, dock contact, or emergency stop?

This is the same buyer logic ui44 already applies to payload, runtime, privacy, repair, and autonomy. A single top-line spec is not enough when the task changes the risk. A robot that is quiet while standing may be noisy when carrying a full laundry basket. A robot that is quiet on carpet may clack on tile. A robot that moves silently may still wake everyone with voice confirmations.

This is also where form factor matters. Hello Robot Stretch 4 is not a humanoid, but it is a useful comparison point because it is a real, available home/workplace mobile manipulator at $29,950, with an 8-hour light-load runtime, 160 cm working height, 45 cm footprint, self-charging, and a 2.5 kg extended / 4 kg retracted payload. A wheeled mobile manipulator may avoid footfall noise entirely, but it still needs motor, wheel, fan, and alert data if it will work around people.

The lesson is not that wheels are automatically quiet or legs are automatically loud. The lesson is that home robotics needs measurement by task.

What ui44 should track next

Noise should become a first-class field for home robots, but only if the data is honest. A future ui44 comparison should separate:

• disclosed noise number;
• dB or dB(A);
• measurement distance;
• tested task;
• floor or surface;
• quiet mode availability;
• max alert volume;
• dock noise;
• whether the claim is manufacturer-tested, independently tested, or reported by third-party coverage.

Until then, the absence of a noise number is itself a signal. A home robot that publishes price, height, payload, battery life, sensors, and AI stack but not sound has left out a spec that directly affects whether people can live with it.

The best reading of 1X NEO's quiet positioning is not "problem solved." It is "the right question is finally on the spec sheet." For home humanoids, that is a real step forward. Now the category needs comparable, task-specific proof.

If you are comparing robots today, use the ui44 compare tool to look at the basics first: status, price, height, weight, runtime, payload, sensors, and use case. Then add one more question that most spec sheets still dodge:

Can this robot do useful work in your home without making the home worse to live in?