# I reverse engineered Deepinder Goyal's Temple

*Published November 16, 2026 · Product & UX Analysis, Indian Startups & Business · by Krishna Veera Vanamali*

Canonical: https://whyvanamali.com/writing/reverse-engineered-deepinder-goyals-temple
Originally posted on LinkedIn: https://www.linkedin.com/posts/whyvanamali_i-reverse-engineered-deepinder-goyals-temple-activity-7527951862919024640-UiWB

Video: https://whyvanamali-audio.vanakrish.workers.dev/video/reverse-engineered-deepinder-goyals-temple.mp4

I reverse engineered Deepinder Goyal's Temple so my laptop can hear me without a microphone.

Not exactly (I don't have the device) but I wanted to test his claim that physical micro-vibrations generated through the skin and skull when speaking can be converted into text. I thought it should be possible at some level with my laptop because it has an accelerometer, a type of motion sensor.

Sound is vibration, so when there is sound, the body of the laptop shakes a little. My question was whether the accelerometer could record that shaking, and whether I could turn the recording back into audio and transcribe it.

Talking at the laptop did nothing. I shouted, I hummed, I even talked while resting my chin on it. The sensor picked up nothing useful. It all comes down to frequency. In my tests, it only picked up a usable signal from vibrations below about 200 Hz, and anything higher got lost in the sensor’s own noise. The frequencies that make speech intelligible (like when we speak) sit much higher, between ~300 and 3,400 Hz, so my voice never showed up in the readings.

So I took a short sentence and slowed it down 40x, which lowers every frequency in it by the same factor. I played that through the laptop’s own speakers with the microphone input off, and let the accelerometer record the vibrations in the laptop’s body in a silent environment. Then I sped the accelerometer’s recording back up 40x and gave it to a speech-to-text model.

It came back with every word right.

But this only worked on slowed audio played from the laptop speakers directly. Which makes sense because Temple sits pressed against your head with what I'd guess is a much faster bone-conduction sensor with higher sensitivity than this. This is also why Temple can't 'hear' people around/near you because the sensor can only detect physical surface vibrations.

## In the video

Krishna on camera, with screen recordings of the experiment, Deepinder Goyal's post about Temple, and on-screen captions. Transcript, lightly edited (repetitions removed):

A few days back, Deepinder Goyal announced Temple again. It's a small device you stick on the side of your head; you double-tap, talk to it, and it captures your thoughts. He made one incredible claim: it has no microphone. So how does it hear you? He says it has motion sensors that read the vibrations your voice makes and turn them into text. I wanted to understand if that can even work, so I tried to recreate it with the one thing I have: my MacBook. My question was simple. If the laptop vibrates when there's sound, can the accelerometer record that vibration, and can I reconstruct audio from it and transcribe it? Turns out it is possible.

This is the sentence I'm using, read by a computer voice: "Temple can hear me without a microphone because it feels the vibrations of my voice." Now I'll run the test: the same sentence, slowed down 40 times, playing out of the laptop speakers. You can hear that low hum; five and a half seconds of speech becomes almost three and a half minutes, all under 200 hertz. While it plays, the accelerometer records the vibration of the laptop body.

Here's what came back: "The Temple can hear me without a microphone because it feels the vibrations of my voice." It got almost every word right, with one extra "the" at the beginning. This transcript came from the accelerometer data; no original audio and no microphone were involved. This is the slowed version the speakers played, and this is the audio reconstructed from the accelerometer, sped back up 40 times. There's a lot of noise, but the sentence is there.

Here's the thing. The laptop already has a motion sensor in it, an accelerometer, the same kind of chip that tells your phone it has been rotated. And sound is essentially vibration: when I talk, I push pressure waves through the air, and every surface they hit vibrates a little, including the body of this laptop.

First I tried the obvious thing. I talked at the laptop. I shouted. I hummed. I even rested my chin on the palm rest and spoke, so my voice would conduct straight into the chassis. The accelerometer picked up nothing I could use. The reason is frequency. The accelerometer samples about 800 times a second and only responds to vibrations below about 200 hertz, but the frequencies that make speech intelligible sit between roughly 300 and 3,400 hertz. Everything that makes my voice understandable to a transcription model is above what this sensor can register.

So, with a bit of help from Claude, I changed the test: if the sensor only registers low frequencies, move the speech down to low frequencies. I took one sentence, about four to five seconds long, and slowed it down about 40 times. Slowing audio lowers every frequency in it by the same factor, so the voice landed between 80 and 160 hertz, inside the band the accelerometer responds to. I played that through the laptop's own speakers and let the accelerometer record the chassis vibrating. Then I sped the recording back up 40 times to restore the original frequencies and passed it to Whisper, an open-source speech-to-text model.

To be clear about what this shows: my laptop cannot listen to me talk. It only worked because I slowed the sentence down 40 times and played it from the laptop's own speakers. But I did it with no extra hardware, on an accelerometer clearly never built for audio, and it figured out the whole sentence.

Temple sits on your head, pressed against the bone, and I'd guess it uses a sensor with a much higher sample rate than my laptop's. Your own voice reaches it by bone conduction, which is far stronger than sound arriving through the air from across the room. That would also explain the other thing Deepinder said: that someone can shout out loud nearby and the device won't pick it up. I don't know what's inside Temple, but after this, recording with no microphone doesn't sound so crazy to me anymore.

---

Next: [India's top 100 AI startups, a market map](https://whyvanamali.com/writing/indias-top-100-ai-startups-market-map)
All posts: https://whyvanamali.com/writing
